|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 0.9993849938499385, |
|
"eval_steps": 500, |
|
"global_step": 1625, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.8514, |
|
"step": 1 |
|
}, |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.3766, |
|
"step": 2 |
|
}, |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.2801, |
|
"step": 3 |
|
}, |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.2422, |
|
"step": 4 |
|
}, |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.3202, |
|
"step": 5 |
|
}, |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.2358, |
|
"step": 6 |
|
}, |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.1181, |
|
"step": 7 |
|
}, |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.1237, |
|
"step": 8 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.2216, |
|
"step": 9 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.1798, |
|
"step": 10 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.2057, |
|
"step": 11 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.427, |
|
"step": 12 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.109, |
|
"step": 13 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.3018, |
|
"step": 14 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.0328, |
|
"step": 15 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.1092, |
|
"step": 16 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.0583, |
|
"step": 17 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.2828, |
|
"step": 18 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.2697, |
|
"step": 19 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.079, |
|
"step": 20 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.1403, |
|
"step": 21 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.1551, |
|
"step": 22 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.1355, |
|
"step": 23 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 0.0, |
|
"loss": 1.0841, |
|
"step": 24 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.274, |
|
"step": 25 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.1752, |
|
"step": 26 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.0604, |
|
"step": 27 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.1368, |
|
"step": 28 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.277, |
|
"step": 29 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 0.8973, |
|
"step": 30 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.2097, |
|
"step": 31 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.055, |
|
"step": 32 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.1259, |
|
"step": 33 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.1009, |
|
"step": 34 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 0.9731, |
|
"step": 35 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.0535, |
|
"step": 36 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.2877, |
|
"step": 37 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.1483, |
|
"step": 38 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.1583, |
|
"step": 39 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 0.0, |
|
"loss": 1.2481, |
|
"step": 40 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.0012, |
|
"step": 41 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.1781, |
|
"step": 42 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.1054, |
|
"step": 43 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.0779, |
|
"step": 44 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.1545, |
|
"step": 45 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.502, |
|
"step": 46 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.2634, |
|
"step": 47 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.2905, |
|
"step": 48 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.1528, |
|
"step": 49 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.3524, |
|
"step": 50 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.1892, |
|
"step": 51 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.2449, |
|
"step": 52 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.075, |
|
"step": 53 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.1034, |
|
"step": 54 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.2607, |
|
"step": 55 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 0.0, |
|
"loss": 1.1986, |
|
"step": 56 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.0669, |
|
"step": 57 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.0169, |
|
"step": 58 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 0.9892, |
|
"step": 59 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.2873, |
|
"step": 60 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.2022, |
|
"step": 61 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.2681, |
|
"step": 62 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.2956, |
|
"step": 63 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.1538, |
|
"step": 64 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.0868, |
|
"step": 65 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.0995, |
|
"step": 66 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.0952, |
|
"step": 67 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.246, |
|
"step": 68 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.2481, |
|
"step": 69 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.1314, |
|
"step": 70 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.4217, |
|
"step": 71 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.1194, |
|
"step": 72 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 0.0, |
|
"loss": 1.1785, |
|
"step": 73 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.2932, |
|
"step": 74 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.0704, |
|
"step": 75 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.1056, |
|
"step": 76 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.1764, |
|
"step": 77 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.0504, |
|
"step": 78 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.2382, |
|
"step": 79 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.0737, |
|
"step": 80 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 0.9174, |
|
"step": 81 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.1491, |
|
"step": 82 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.1204, |
|
"step": 83 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.0855, |
|
"step": 84 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.2597, |
|
"step": 85 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.2517, |
|
"step": 86 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.3423, |
|
"step": 87 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.2276, |
|
"step": 88 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 0.0, |
|
"loss": 1.0958, |
|
"step": 89 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.2448, |
|
"step": 90 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.0191, |
|
"step": 91 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.2652, |
|
"step": 92 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.111, |
|
"step": 93 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.2031, |
|
"step": 94 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.464, |
|
"step": 95 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.1549, |
|
"step": 96 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.262, |
|
"step": 97 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.1474, |
|
"step": 98 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.199, |
|
"step": 99 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.209, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.2007, |
|
"step": 101 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.3153, |
|
"step": 102 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.1739, |
|
"step": 103 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.2299, |
|
"step": 104 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 0.0, |
|
"loss": 1.2235, |
|
"step": 105 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.1972, |
|
"step": 106 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.1649, |
|
"step": 107 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.2151, |
|
"step": 108 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.2605, |
|
"step": 109 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.1389, |
|
"step": 110 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.033, |
|
"step": 111 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.2221, |
|
"step": 112 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.2068, |
|
"step": 113 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.0533, |
|
"step": 114 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.1951, |
|
"step": 115 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.3248, |
|
"step": 116 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.0856, |
|
"step": 117 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.0907, |
|
"step": 118 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.3091, |
|
"step": 119 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.1677, |
|
"step": 120 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 0.0, |
|
"loss": 1.0901, |
|
"step": 121 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.1494, |
|
"step": 122 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.1736, |
|
"step": 123 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.1291, |
|
"step": 124 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.2865, |
|
"step": 125 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.2271, |
|
"step": 126 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.168, |
|
"step": 127 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.1429, |
|
"step": 128 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.0094, |
|
"step": 129 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.0979, |
|
"step": 130 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.1221, |
|
"step": 131 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.3071, |
|
"step": 132 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.0912, |
|
"step": 133 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.1448, |
|
"step": 134 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.11, |
|
"step": 135 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.0182, |
|
"step": 136 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.0496, |
|
"step": 137 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 0.0, |
|
"loss": 1.0395, |
|
"step": 138 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.2406, |
|
"step": 139 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.1877, |
|
"step": 140 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.2488, |
|
"step": 141 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.2292, |
|
"step": 142 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.1267, |
|
"step": 143 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.2374, |
|
"step": 144 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 0.9793, |
|
"step": 145 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.312, |
|
"step": 146 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.3403, |
|
"step": 147 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.17, |
|
"step": 148 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.3125, |
|
"step": 149 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.2386, |
|
"step": 150 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.1924, |
|
"step": 151 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.1742, |
|
"step": 152 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.0657, |
|
"step": 153 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 0.0, |
|
"loss": 1.2138, |
|
"step": 154 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.1157, |
|
"step": 155 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.3357, |
|
"step": 156 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.1364, |
|
"step": 157 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.0077, |
|
"step": 158 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.1201, |
|
"step": 159 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.1124, |
|
"step": 160 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.3079, |
|
"step": 161 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.2359, |
|
"step": 162 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.2958, |
|
"step": 163 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.2084, |
|
"step": 164 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.2024, |
|
"step": 165 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.1337, |
|
"step": 166 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.0667, |
|
"step": 167 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.1691, |
|
"step": 168 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.1014, |
|
"step": 169 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 0.0, |
|
"loss": 1.0879, |
|
"step": 170 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.2519, |
|
"step": 171 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.2245, |
|
"step": 172 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.1155, |
|
"step": 173 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.125, |
|
"step": 174 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.0487, |
|
"step": 175 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.1436, |
|
"step": 176 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.2218, |
|
"step": 177 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.0585, |
|
"step": 178 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.2107, |
|
"step": 179 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.2978, |
|
"step": 180 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.2423, |
|
"step": 181 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.2064, |
|
"step": 182 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.1567, |
|
"step": 183 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.2222, |
|
"step": 184 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.1356, |
|
"step": 185 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 0.0, |
|
"loss": 1.0913, |
|
"step": 186 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 0.993, |
|
"step": 187 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.0263, |
|
"step": 188 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1463, |
|
"step": 189 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1321, |
|
"step": 190 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1875, |
|
"step": 191 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.2995, |
|
"step": 192 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1525, |
|
"step": 193 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1285, |
|
"step": 194 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1812, |
|
"step": 195 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1247, |
|
"step": 196 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1087, |
|
"step": 197 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1764, |
|
"step": 198 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1186, |
|
"step": 199 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.1892, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.2106, |
|
"step": 201 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.2797, |
|
"step": 202 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 0.0, |
|
"loss": 1.3146, |
|
"step": 203 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.1978, |
|
"step": 204 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.0686, |
|
"step": 205 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.2571, |
|
"step": 206 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.2779, |
|
"step": 207 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.1508, |
|
"step": 208 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.4313, |
|
"step": 209 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.0895, |
|
"step": 210 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.0953, |
|
"step": 211 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.2328, |
|
"step": 212 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.0915, |
|
"step": 213 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.1392, |
|
"step": 214 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.1408, |
|
"step": 215 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.1229, |
|
"step": 216 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.1227, |
|
"step": 217 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.0799, |
|
"step": 218 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 0.0, |
|
"loss": 1.2121, |
|
"step": 219 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 0.9775, |
|
"step": 220 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.1368, |
|
"step": 221 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.1674, |
|
"step": 222 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.3371, |
|
"step": 223 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.1432, |
|
"step": 224 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.1908, |
|
"step": 225 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.1949, |
|
"step": 226 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.1881, |
|
"step": 227 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.229, |
|
"step": 228 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.2421, |
|
"step": 229 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.2449, |
|
"step": 230 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.2099, |
|
"step": 231 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.0306, |
|
"step": 232 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.0804, |
|
"step": 233 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.1436, |
|
"step": 234 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 0.0, |
|
"loss": 1.0791, |
|
"step": 235 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.226, |
|
"step": 236 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.1442, |
|
"step": 237 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.2, |
|
"step": 238 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.2428, |
|
"step": 239 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.0922, |
|
"step": 240 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.0125, |
|
"step": 241 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.1576, |
|
"step": 242 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.1524, |
|
"step": 243 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.2514, |
|
"step": 244 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.1804, |
|
"step": 245 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.2935, |
|
"step": 246 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.1409, |
|
"step": 247 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.2423, |
|
"step": 248 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.2861, |
|
"step": 249 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.1522, |
|
"step": 250 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.0477, |
|
"step": 251 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 0.0, |
|
"loss": 1.1086, |
|
"step": 252 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.1966, |
|
"step": 253 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.0852, |
|
"step": 254 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.099, |
|
"step": 255 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.2005, |
|
"step": 256 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.225, |
|
"step": 257 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.2122, |
|
"step": 258 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.1485, |
|
"step": 259 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.1563, |
|
"step": 260 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.0895, |
|
"step": 261 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.1855, |
|
"step": 262 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.2767, |
|
"step": 263 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.2954, |
|
"step": 264 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.2078, |
|
"step": 265 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.0166, |
|
"step": 266 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.4389, |
|
"step": 267 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 0.0, |
|
"loss": 1.152, |
|
"step": 268 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.1859, |
|
"step": 269 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.1471, |
|
"step": 270 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 0.9637, |
|
"step": 271 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.0884, |
|
"step": 272 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.1041, |
|
"step": 273 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.1352, |
|
"step": 274 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.2467, |
|
"step": 275 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.163, |
|
"step": 276 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.2596, |
|
"step": 277 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.2577, |
|
"step": 278 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.2647, |
|
"step": 279 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.1175, |
|
"step": 280 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.2303, |
|
"step": 281 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.1619, |
|
"step": 282 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.1023, |
|
"step": 283 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 0.0, |
|
"loss": 1.1853, |
|
"step": 284 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.3678, |
|
"step": 285 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.3368, |
|
"step": 286 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.1719, |
|
"step": 287 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.0129, |
|
"step": 288 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.1908, |
|
"step": 289 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.1789, |
|
"step": 290 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.1391, |
|
"step": 291 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.2972, |
|
"step": 292 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.1541, |
|
"step": 293 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.1853, |
|
"step": 294 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.213, |
|
"step": 295 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.1132, |
|
"step": 296 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.2317, |
|
"step": 297 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.1325, |
|
"step": 298 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.2584, |
|
"step": 299 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.14, |
|
"step": 300 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.0595, |
|
"step": 301 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.2193, |
|
"step": 302 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.3712, |
|
"step": 303 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.2972, |
|
"step": 304 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.2931, |
|
"step": 305 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.2775, |
|
"step": 306 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.1124, |
|
"step": 307 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.1335, |
|
"step": 308 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.0518, |
|
"step": 309 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.1242, |
|
"step": 310 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.165, |
|
"step": 311 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.1306, |
|
"step": 312 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.2977, |
|
"step": 313 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.0734, |
|
"step": 314 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.2207, |
|
"step": 315 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.1507, |
|
"step": 316 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 0.0, |
|
"loss": 1.1302, |
|
"step": 317 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.2222, |
|
"step": 318 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.0854, |
|
"step": 319 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.2083, |
|
"step": 320 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.2257, |
|
"step": 321 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.1191, |
|
"step": 322 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.1876, |
|
"step": 323 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.2627, |
|
"step": 324 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.1374, |
|
"step": 325 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.2955, |
|
"step": 326 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.1682, |
|
"step": 327 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.0484, |
|
"step": 328 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.1175, |
|
"step": 329 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.1659, |
|
"step": 330 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.2698, |
|
"step": 331 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.2167, |
|
"step": 332 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 0.0, |
|
"loss": 1.1599, |
|
"step": 333 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.1531, |
|
"step": 334 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.1669, |
|
"step": 335 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.2543, |
|
"step": 336 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.25, |
|
"step": 337 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.2254, |
|
"step": 338 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.1986, |
|
"step": 339 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.1263, |
|
"step": 340 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.1416, |
|
"step": 341 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.3271, |
|
"step": 342 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.0922, |
|
"step": 343 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.2447, |
|
"step": 344 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.1202, |
|
"step": 345 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.218, |
|
"step": 346 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.2542, |
|
"step": 347 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.1516, |
|
"step": 348 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 0.0, |
|
"loss": 1.162, |
|
"step": 349 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.15, |
|
"step": 350 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.1851, |
|
"step": 351 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.2972, |
|
"step": 352 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.0101, |
|
"step": 353 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.2194, |
|
"step": 354 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.2295, |
|
"step": 355 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.1917, |
|
"step": 356 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.2543, |
|
"step": 357 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.2661, |
|
"step": 358 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.0695, |
|
"step": 359 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.0384, |
|
"step": 360 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.0885, |
|
"step": 361 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.0681, |
|
"step": 362 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.3068, |
|
"step": 363 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.1518, |
|
"step": 364 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 0.0, |
|
"loss": 1.1383, |
|
"step": 365 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.3121, |
|
"step": 366 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2708, |
|
"step": 367 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.3244, |
|
"step": 368 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2719, |
|
"step": 369 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.226, |
|
"step": 370 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.0039, |
|
"step": 371 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2458, |
|
"step": 372 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2167, |
|
"step": 373 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2005, |
|
"step": 374 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.1664, |
|
"step": 375 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.228, |
|
"step": 376 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2587, |
|
"step": 377 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.0361, |
|
"step": 378 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2021, |
|
"step": 379 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.1116, |
|
"step": 380 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2726, |
|
"step": 381 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 0.0, |
|
"loss": 1.2745, |
|
"step": 382 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.2178, |
|
"step": 383 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.262, |
|
"step": 384 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.2336, |
|
"step": 385 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.1445, |
|
"step": 386 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.221, |
|
"step": 387 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.2842, |
|
"step": 388 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.0023, |
|
"step": 389 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.1859, |
|
"step": 390 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.0835, |
|
"step": 391 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.2852, |
|
"step": 392 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.2012, |
|
"step": 393 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.2997, |
|
"step": 394 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.1679, |
|
"step": 395 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.1969, |
|
"step": 396 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.2461, |
|
"step": 397 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 0.0, |
|
"loss": 1.1386, |
|
"step": 398 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.0386, |
|
"step": 399 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.2279, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 0.9942, |
|
"step": 401 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.3307, |
|
"step": 402 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.2367, |
|
"step": 403 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.2916, |
|
"step": 404 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.2725, |
|
"step": 405 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.0696, |
|
"step": 406 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.0929, |
|
"step": 407 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.3157, |
|
"step": 408 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.194, |
|
"step": 409 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.1562, |
|
"step": 410 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.0008, |
|
"step": 411 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.182, |
|
"step": 412 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.2729, |
|
"step": 413 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 0.0, |
|
"loss": 1.0641, |
|
"step": 414 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.0112, |
|
"step": 415 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.2515, |
|
"step": 416 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.0657, |
|
"step": 417 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.1887, |
|
"step": 418 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.137, |
|
"step": 419 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.2397, |
|
"step": 420 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.151, |
|
"step": 421 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.317, |
|
"step": 422 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.271, |
|
"step": 423 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.1251, |
|
"step": 424 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.1345, |
|
"step": 425 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.0799, |
|
"step": 426 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.0626, |
|
"step": 427 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.0366, |
|
"step": 428 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.2921, |
|
"step": 429 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 0.0, |
|
"loss": 1.2031, |
|
"step": 430 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.1343, |
|
"step": 431 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.1207, |
|
"step": 432 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.3368, |
|
"step": 433 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.317, |
|
"step": 434 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.3293, |
|
"step": 435 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.1569, |
|
"step": 436 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.075, |
|
"step": 437 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.314, |
|
"step": 438 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.3229, |
|
"step": 439 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.2344, |
|
"step": 440 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.0496, |
|
"step": 441 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.3267, |
|
"step": 442 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.4823, |
|
"step": 443 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.2057, |
|
"step": 444 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.0277, |
|
"step": 445 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.2191, |
|
"step": 446 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 0.0, |
|
"loss": 1.1777, |
|
"step": 447 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.2837, |
|
"step": 448 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.1844, |
|
"step": 449 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.2944, |
|
"step": 450 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.1631, |
|
"step": 451 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.2179, |
|
"step": 452 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.2619, |
|
"step": 453 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.1457, |
|
"step": 454 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.1966, |
|
"step": 455 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.122, |
|
"step": 456 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.2157, |
|
"step": 457 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.2407, |
|
"step": 458 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.1318, |
|
"step": 459 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.0884, |
|
"step": 460 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.3816, |
|
"step": 461 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.2666, |
|
"step": 462 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 0.0, |
|
"loss": 1.1258, |
|
"step": 463 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.2358, |
|
"step": 464 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.3154, |
|
"step": 465 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.2444, |
|
"step": 466 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.0764, |
|
"step": 467 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.3218, |
|
"step": 468 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.1883, |
|
"step": 469 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 0.957, |
|
"step": 470 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.2419, |
|
"step": 471 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.2016, |
|
"step": 472 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.0284, |
|
"step": 473 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.1063, |
|
"step": 474 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.2774, |
|
"step": 475 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.0947, |
|
"step": 476 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.2229, |
|
"step": 477 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.2469, |
|
"step": 478 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 0.0, |
|
"loss": 1.0293, |
|
"step": 479 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.1542, |
|
"step": 480 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.2603, |
|
"step": 481 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.3458, |
|
"step": 482 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.1332, |
|
"step": 483 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.0063, |
|
"step": 484 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.2083, |
|
"step": 485 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.3951, |
|
"step": 486 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.148, |
|
"step": 487 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.2005, |
|
"step": 488 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 0.9837, |
|
"step": 489 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.0827, |
|
"step": 490 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.0053, |
|
"step": 491 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.0839, |
|
"step": 492 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.2075, |
|
"step": 493 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.29, |
|
"step": 494 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 0.0, |
|
"loss": 1.19, |
|
"step": 495 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.3059, |
|
"step": 496 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.2826, |
|
"step": 497 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.1604, |
|
"step": 498 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.1383, |
|
"step": 499 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.2319, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.1851, |
|
"step": 501 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.147, |
|
"step": 502 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.2363, |
|
"step": 503 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 0.973, |
|
"step": 504 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.145, |
|
"step": 505 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.2719, |
|
"step": 506 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.284, |
|
"step": 507 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.1082, |
|
"step": 508 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.247, |
|
"step": 509 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.0236, |
|
"step": 510 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.1562, |
|
"step": 511 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 0.0, |
|
"loss": 1.0443, |
|
"step": 512 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.1503, |
|
"step": 513 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.1273, |
|
"step": 514 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.0362, |
|
"step": 515 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.2111, |
|
"step": 516 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.3323, |
|
"step": 517 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.1254, |
|
"step": 518 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.0719, |
|
"step": 519 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.103, |
|
"step": 520 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.129, |
|
"step": 521 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.2561, |
|
"step": 522 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.2864, |
|
"step": 523 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.124, |
|
"step": 524 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.0491, |
|
"step": 525 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.0742, |
|
"step": 526 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.0562, |
|
"step": 527 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 0.0, |
|
"loss": 1.3065, |
|
"step": 528 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.198, |
|
"step": 529 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.0974, |
|
"step": 530 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.0685, |
|
"step": 531 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.1218, |
|
"step": 532 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.2546, |
|
"step": 533 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.2117, |
|
"step": 534 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.1666, |
|
"step": 535 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.0974, |
|
"step": 536 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.2456, |
|
"step": 537 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.264, |
|
"step": 538 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.288, |
|
"step": 539 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.1839, |
|
"step": 540 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.2489, |
|
"step": 541 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.2093, |
|
"step": 542 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.2189, |
|
"step": 543 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 0.0, |
|
"loss": 1.2868, |
|
"step": 544 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.2873, |
|
"step": 545 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.1874, |
|
"step": 546 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.215, |
|
"step": 547 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.3476, |
|
"step": 548 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.3138, |
|
"step": 549 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.0758, |
|
"step": 550 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.2278, |
|
"step": 551 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.2149, |
|
"step": 552 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.0834, |
|
"step": 553 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.164, |
|
"step": 554 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.1737, |
|
"step": 555 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.2099, |
|
"step": 556 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.0118, |
|
"step": 557 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.1571, |
|
"step": 558 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.1784, |
|
"step": 559 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 0.0, |
|
"loss": 1.1267, |
|
"step": 560 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.2353, |
|
"step": 561 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.0943, |
|
"step": 562 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.1568, |
|
"step": 563 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.2189, |
|
"step": 564 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.293, |
|
"step": 565 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.2385, |
|
"step": 566 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.4454, |
|
"step": 567 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.2165, |
|
"step": 568 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.3574, |
|
"step": 569 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.1736, |
|
"step": 570 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.1036, |
|
"step": 571 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.2698, |
|
"step": 572 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.33, |
|
"step": 573 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.2371, |
|
"step": 574 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.1814, |
|
"step": 575 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.2049, |
|
"step": 576 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 0.0, |
|
"loss": 1.1452, |
|
"step": 577 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.141, |
|
"step": 578 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.3718, |
|
"step": 579 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.2162, |
|
"step": 580 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.1703, |
|
"step": 581 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.1095, |
|
"step": 582 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.256, |
|
"step": 583 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.2563, |
|
"step": 584 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.3619, |
|
"step": 585 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.1999, |
|
"step": 586 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.008, |
|
"step": 587 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.1217, |
|
"step": 588 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.2443, |
|
"step": 589 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.2102, |
|
"step": 590 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.0754, |
|
"step": 591 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.195, |
|
"step": 592 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 0.0, |
|
"loss": 1.0842, |
|
"step": 593 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.1055, |
|
"step": 594 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.4119, |
|
"step": 595 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.1949, |
|
"step": 596 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.1664, |
|
"step": 597 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.279, |
|
"step": 598 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.5431, |
|
"step": 599 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.2273, |
|
"step": 600 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.2985, |
|
"step": 601 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.21, |
|
"step": 602 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.2351, |
|
"step": 603 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.0486, |
|
"step": 604 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.1116, |
|
"step": 605 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.406, |
|
"step": 606 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.1564, |
|
"step": 607 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.1468, |
|
"step": 608 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 0.0, |
|
"loss": 1.3805, |
|
"step": 609 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.1466, |
|
"step": 610 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.219, |
|
"step": 611 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.314, |
|
"step": 612 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.2127, |
|
"step": 613 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.252, |
|
"step": 614 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.1486, |
|
"step": 615 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.3345, |
|
"step": 616 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.2528, |
|
"step": 617 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.0423, |
|
"step": 618 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.2258, |
|
"step": 619 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.1283, |
|
"step": 620 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.2521, |
|
"step": 621 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.1776, |
|
"step": 622 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.1348, |
|
"step": 623 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.2751, |
|
"step": 624 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.1982, |
|
"step": 625 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 0.0, |
|
"loss": 1.1793, |
|
"step": 626 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.1424, |
|
"step": 627 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.2003, |
|
"step": 628 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.2144, |
|
"step": 629 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.1988, |
|
"step": 630 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.301, |
|
"step": 631 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.291, |
|
"step": 632 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.0967, |
|
"step": 633 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.2132, |
|
"step": 634 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.0812, |
|
"step": 635 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.187, |
|
"step": 636 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.1506, |
|
"step": 637 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.2741, |
|
"step": 638 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.2655, |
|
"step": 639 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.1616, |
|
"step": 640 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.1189, |
|
"step": 641 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 0.0, |
|
"loss": 1.0739, |
|
"step": 642 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.0954, |
|
"step": 643 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.2136, |
|
"step": 644 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.1488, |
|
"step": 645 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.1675, |
|
"step": 646 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.0939, |
|
"step": 647 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.2926, |
|
"step": 648 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.2557, |
|
"step": 649 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.2146, |
|
"step": 650 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.3579, |
|
"step": 651 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.1438, |
|
"step": 652 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 0.8767, |
|
"step": 653 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.2159, |
|
"step": 654 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.2193, |
|
"step": 655 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.21, |
|
"step": 656 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.1849, |
|
"step": 657 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 0.0, |
|
"loss": 1.2749, |
|
"step": 658 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.0678, |
|
"step": 659 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.2589, |
|
"step": 660 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.223, |
|
"step": 661 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.303, |
|
"step": 662 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.3903, |
|
"step": 663 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.112, |
|
"step": 664 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.0804, |
|
"step": 665 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.1152, |
|
"step": 666 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.1561, |
|
"step": 667 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 0.9647, |
|
"step": 668 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.0911, |
|
"step": 669 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.2259, |
|
"step": 670 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.3306, |
|
"step": 671 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.2043, |
|
"step": 672 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.0814, |
|
"step": 673 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 0.0, |
|
"loss": 1.2561, |
|
"step": 674 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1661, |
|
"step": 675 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1418, |
|
"step": 676 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1622, |
|
"step": 677 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.0463, |
|
"step": 678 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.159, |
|
"step": 679 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1525, |
|
"step": 680 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1943, |
|
"step": 681 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1439, |
|
"step": 682 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1859, |
|
"step": 683 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1995, |
|
"step": 684 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.2256, |
|
"step": 685 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.3397, |
|
"step": 686 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1554, |
|
"step": 687 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1561, |
|
"step": 688 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.1196, |
|
"step": 689 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.342, |
|
"step": 690 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 0.0, |
|
"loss": 1.2147, |
|
"step": 691 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.248, |
|
"step": 692 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.2844, |
|
"step": 693 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.2188, |
|
"step": 694 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.1, |
|
"step": 695 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.2885, |
|
"step": 696 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.1576, |
|
"step": 697 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.217, |
|
"step": 698 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.2081, |
|
"step": 699 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.3212, |
|
"step": 700 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.0601, |
|
"step": 701 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.1283, |
|
"step": 702 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.1386, |
|
"step": 703 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.1471, |
|
"step": 704 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.0889, |
|
"step": 705 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.1336, |
|
"step": 706 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 0.0, |
|
"loss": 1.2088, |
|
"step": 707 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.2842, |
|
"step": 708 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.4298, |
|
"step": 709 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.1989, |
|
"step": 710 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.1678, |
|
"step": 711 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.2185, |
|
"step": 712 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.2916, |
|
"step": 713 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.2067, |
|
"step": 714 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.3053, |
|
"step": 715 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.2216, |
|
"step": 716 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 0.9626, |
|
"step": 717 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.1888, |
|
"step": 718 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.3075, |
|
"step": 719 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.2199, |
|
"step": 720 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.2292, |
|
"step": 721 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.244, |
|
"step": 722 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 0.0, |
|
"loss": 1.3725, |
|
"step": 723 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.2534, |
|
"step": 724 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.2953, |
|
"step": 725 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.3018, |
|
"step": 726 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.2467, |
|
"step": 727 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.2395, |
|
"step": 728 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.3388, |
|
"step": 729 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.1676, |
|
"step": 730 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.188, |
|
"step": 731 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.2391, |
|
"step": 732 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.1566, |
|
"step": 733 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.0787, |
|
"step": 734 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.2589, |
|
"step": 735 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.3721, |
|
"step": 736 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.204, |
|
"step": 737 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.2565, |
|
"step": 738 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 0.0, |
|
"loss": 1.1522, |
|
"step": 739 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.2579, |
|
"step": 740 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.1836, |
|
"step": 741 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.3028, |
|
"step": 742 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.1114, |
|
"step": 743 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.2547, |
|
"step": 744 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.4102, |
|
"step": 745 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.2813, |
|
"step": 746 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.1316, |
|
"step": 747 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.3437, |
|
"step": 748 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.3192, |
|
"step": 749 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.2661, |
|
"step": 750 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.2679, |
|
"step": 751 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.2377, |
|
"step": 752 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.059, |
|
"step": 753 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.0908, |
|
"step": 754 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.089, |
|
"step": 755 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 0.0, |
|
"loss": 1.2857, |
|
"step": 756 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.3246, |
|
"step": 757 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.0471, |
|
"step": 758 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2985, |
|
"step": 759 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2331, |
|
"step": 760 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2599, |
|
"step": 761 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.3606, |
|
"step": 762 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.1897, |
|
"step": 763 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2757, |
|
"step": 764 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.1906, |
|
"step": 765 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2116, |
|
"step": 766 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2648, |
|
"step": 767 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.0669, |
|
"step": 768 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2081, |
|
"step": 769 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.1925, |
|
"step": 770 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2157, |
|
"step": 771 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 0.0, |
|
"loss": 1.2894, |
|
"step": 772 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.3251, |
|
"step": 773 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.0867, |
|
"step": 774 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.0244, |
|
"step": 775 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.2809, |
|
"step": 776 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.3123, |
|
"step": 777 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.2181, |
|
"step": 778 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.1283, |
|
"step": 779 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.2599, |
|
"step": 780 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.1679, |
|
"step": 781 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.2259, |
|
"step": 782 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 0.9506, |
|
"step": 783 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.3386, |
|
"step": 784 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.0913, |
|
"step": 785 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.1752, |
|
"step": 786 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.1977, |
|
"step": 787 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 0.0, |
|
"loss": 1.121, |
|
"step": 788 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.2292, |
|
"step": 789 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.3819, |
|
"step": 790 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.3347, |
|
"step": 791 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.273, |
|
"step": 792 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.2287, |
|
"step": 793 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.2191, |
|
"step": 794 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.2189, |
|
"step": 795 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.3406, |
|
"step": 796 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.2304, |
|
"step": 797 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.2762, |
|
"step": 798 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.2743, |
|
"step": 799 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.0879, |
|
"step": 800 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.3689, |
|
"step": 801 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.224, |
|
"step": 802 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.269, |
|
"step": 803 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 0.0, |
|
"loss": 1.3692, |
|
"step": 804 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.2675, |
|
"step": 805 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.1835, |
|
"step": 806 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.1214, |
|
"step": 807 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.1853, |
|
"step": 808 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.1387, |
|
"step": 809 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.2977, |
|
"step": 810 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.273, |
|
"step": 811 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.0218, |
|
"step": 812 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.3124, |
|
"step": 813 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.2274, |
|
"step": 814 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.1786, |
|
"step": 815 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.1469, |
|
"step": 816 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.0348, |
|
"step": 817 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.2956, |
|
"step": 818 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.0968, |
|
"step": 819 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.07, |
|
"step": 820 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 0.0, |
|
"loss": 1.281, |
|
"step": 821 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.215, |
|
"step": 822 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.2211, |
|
"step": 823 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.2618, |
|
"step": 824 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.0732, |
|
"step": 825 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.2336, |
|
"step": 826 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.4517, |
|
"step": 827 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.3503, |
|
"step": 828 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 0.9953, |
|
"step": 829 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.1674, |
|
"step": 830 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.4218, |
|
"step": 831 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.175, |
|
"step": 832 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.1334, |
|
"step": 833 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.363, |
|
"step": 834 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.1654, |
|
"step": 835 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.3313, |
|
"step": 836 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 0.0, |
|
"loss": 1.1082, |
|
"step": 837 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.1463, |
|
"step": 838 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.1582, |
|
"step": 839 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.0075, |
|
"step": 840 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.1752, |
|
"step": 841 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.3805, |
|
"step": 842 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.1994, |
|
"step": 843 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.2024, |
|
"step": 844 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.1519, |
|
"step": 845 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.1381, |
|
"step": 846 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.2771, |
|
"step": 847 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.2906, |
|
"step": 848 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.142, |
|
"step": 849 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.3412, |
|
"step": 850 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.2234, |
|
"step": 851 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.1613, |
|
"step": 852 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 0.0, |
|
"loss": 1.2509, |
|
"step": 853 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.1051, |
|
"step": 854 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.4245, |
|
"step": 855 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.2712, |
|
"step": 856 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.0507, |
|
"step": 857 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.2684, |
|
"step": 858 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.3287, |
|
"step": 859 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.3292, |
|
"step": 860 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.2344, |
|
"step": 861 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.0963, |
|
"step": 862 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.1603, |
|
"step": 863 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.1606, |
|
"step": 864 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.2977, |
|
"step": 865 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.1713, |
|
"step": 866 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.1938, |
|
"step": 867 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.1532, |
|
"step": 868 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 0.0, |
|
"loss": 1.1743, |
|
"step": 869 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.3014, |
|
"step": 870 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.1836, |
|
"step": 871 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2389, |
|
"step": 872 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.1897, |
|
"step": 873 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.1402, |
|
"step": 874 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.3913, |
|
"step": 875 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.0087, |
|
"step": 876 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2574, |
|
"step": 877 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2236, |
|
"step": 878 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.3749, |
|
"step": 879 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2463, |
|
"step": 880 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.3072, |
|
"step": 881 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2939, |
|
"step": 882 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2467, |
|
"step": 883 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2918, |
|
"step": 884 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2048, |
|
"step": 885 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 0.0, |
|
"loss": 1.2277, |
|
"step": 886 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.2103, |
|
"step": 887 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.0845, |
|
"step": 888 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.1831, |
|
"step": 889 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.0383, |
|
"step": 890 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.3131, |
|
"step": 891 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.0431, |
|
"step": 892 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.15, |
|
"step": 893 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.1155, |
|
"step": 894 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.2174, |
|
"step": 895 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.2442, |
|
"step": 896 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.3055, |
|
"step": 897 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.2999, |
|
"step": 898 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.2529, |
|
"step": 899 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.3141, |
|
"step": 900 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.1373, |
|
"step": 901 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 0.0, |
|
"loss": 1.23, |
|
"step": 902 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.1617, |
|
"step": 903 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.3045, |
|
"step": 904 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.1918, |
|
"step": 905 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.105, |
|
"step": 906 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.2758, |
|
"step": 907 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.0411, |
|
"step": 908 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.3009, |
|
"step": 909 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.2011, |
|
"step": 910 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.2117, |
|
"step": 911 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.1676, |
|
"step": 912 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.1235, |
|
"step": 913 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.1895, |
|
"step": 914 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.25, |
|
"step": 915 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.3228, |
|
"step": 916 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.1286, |
|
"step": 917 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 0.0, |
|
"loss": 1.2222, |
|
"step": 918 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.3073, |
|
"step": 919 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2005, |
|
"step": 920 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2677, |
|
"step": 921 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.1573, |
|
"step": 922 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.1313, |
|
"step": 923 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2784, |
|
"step": 924 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2261, |
|
"step": 925 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2173, |
|
"step": 926 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2955, |
|
"step": 927 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.3698, |
|
"step": 928 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.1876, |
|
"step": 929 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2437, |
|
"step": 930 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2966, |
|
"step": 931 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2991, |
|
"step": 932 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.1593, |
|
"step": 933 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 0.0, |
|
"loss": 1.2808, |
|
"step": 934 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.3065, |
|
"step": 935 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.1073, |
|
"step": 936 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.2826, |
|
"step": 937 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.1483, |
|
"step": 938 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.3711, |
|
"step": 939 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.3202, |
|
"step": 940 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.2351, |
|
"step": 941 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.3704, |
|
"step": 942 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.3142, |
|
"step": 943 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.2245, |
|
"step": 944 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.0907, |
|
"step": 945 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.1042, |
|
"step": 946 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.2276, |
|
"step": 947 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.3296, |
|
"step": 948 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.2728, |
|
"step": 949 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.1725, |
|
"step": 950 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 0.0, |
|
"loss": 1.1255, |
|
"step": 951 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.0806, |
|
"step": 952 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.2586, |
|
"step": 953 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.226, |
|
"step": 954 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.1242, |
|
"step": 955 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.3394, |
|
"step": 956 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.3268, |
|
"step": 957 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.1566, |
|
"step": 958 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.1444, |
|
"step": 959 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.219, |
|
"step": 960 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.3982, |
|
"step": 961 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.1682, |
|
"step": 962 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.2625, |
|
"step": 963 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.2147, |
|
"step": 964 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.3057, |
|
"step": 965 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.1913, |
|
"step": 966 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 0.0, |
|
"loss": 1.1672, |
|
"step": 967 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.2074, |
|
"step": 968 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.3113, |
|
"step": 969 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.2484, |
|
"step": 970 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.181, |
|
"step": 971 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.149, |
|
"step": 972 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.249, |
|
"step": 973 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.0473, |
|
"step": 974 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.1535, |
|
"step": 975 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.0658, |
|
"step": 976 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.1389, |
|
"step": 977 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.2835, |
|
"step": 978 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.1635, |
|
"step": 979 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.3368, |
|
"step": 980 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.2689, |
|
"step": 981 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.2681, |
|
"step": 982 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 0.0, |
|
"loss": 1.3394, |
|
"step": 983 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.2268, |
|
"step": 984 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.2804, |
|
"step": 985 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.2665, |
|
"step": 986 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.4186, |
|
"step": 987 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.3615, |
|
"step": 988 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.202, |
|
"step": 989 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.3495, |
|
"step": 990 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.0304, |
|
"step": 991 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.1924, |
|
"step": 992 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.2607, |
|
"step": 993 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.1576, |
|
"step": 994 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.3292, |
|
"step": 995 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.2646, |
|
"step": 996 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.1819, |
|
"step": 997 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.3548, |
|
"step": 998 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 0.0, |
|
"loss": 1.1643, |
|
"step": 999 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.3354, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2547, |
|
"step": 1001 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2619, |
|
"step": 1002 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2636, |
|
"step": 1003 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2415, |
|
"step": 1004 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.3826, |
|
"step": 1005 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.397, |
|
"step": 1006 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.0364, |
|
"step": 1007 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.1695, |
|
"step": 1008 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2975, |
|
"step": 1009 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.3461, |
|
"step": 1010 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2681, |
|
"step": 1011 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2269, |
|
"step": 1012 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2897, |
|
"step": 1013 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.1617, |
|
"step": 1014 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2712, |
|
"step": 1015 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 0.0, |
|
"loss": 1.2343, |
|
"step": 1016 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.3168, |
|
"step": 1017 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2613, |
|
"step": 1018 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2204, |
|
"step": 1019 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2809, |
|
"step": 1020 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2635, |
|
"step": 1021 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2359, |
|
"step": 1022 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2859, |
|
"step": 1023 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.3221, |
|
"step": 1024 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2189, |
|
"step": 1025 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2499, |
|
"step": 1026 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.327, |
|
"step": 1027 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2824, |
|
"step": 1028 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2362, |
|
"step": 1029 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.1689, |
|
"step": 1030 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.2859, |
|
"step": 1031 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 0.0, |
|
"loss": 1.3683, |
|
"step": 1032 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.3986, |
|
"step": 1033 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.357, |
|
"step": 1034 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.2196, |
|
"step": 1035 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.2985, |
|
"step": 1036 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.1761, |
|
"step": 1037 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.2733, |
|
"step": 1038 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.2629, |
|
"step": 1039 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.2809, |
|
"step": 1040 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.2842, |
|
"step": 1041 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.2492, |
|
"step": 1042 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.2589, |
|
"step": 1043 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.1575, |
|
"step": 1044 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.4573, |
|
"step": 1045 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.218, |
|
"step": 1046 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.114, |
|
"step": 1047 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 0.0, |
|
"loss": 1.1777, |
|
"step": 1048 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.2042, |
|
"step": 1049 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.317, |
|
"step": 1050 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.315, |
|
"step": 1051 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.2666, |
|
"step": 1052 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.1635, |
|
"step": 1053 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.2266, |
|
"step": 1054 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.2177, |
|
"step": 1055 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.2674, |
|
"step": 1056 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.3347, |
|
"step": 1057 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.278, |
|
"step": 1058 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.3165, |
|
"step": 1059 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.196, |
|
"step": 1060 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.2637, |
|
"step": 1061 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.2408, |
|
"step": 1062 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.4121, |
|
"step": 1063 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.4001, |
|
"step": 1064 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 0.0, |
|
"loss": 1.3082, |
|
"step": 1065 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.0597, |
|
"step": 1066 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.3155, |
|
"step": 1067 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.4972, |
|
"step": 1068 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.2948, |
|
"step": 1069 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.3119, |
|
"step": 1070 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.187, |
|
"step": 1071 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.3296, |
|
"step": 1072 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.2621, |
|
"step": 1073 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.2879, |
|
"step": 1074 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.3257, |
|
"step": 1075 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.3339, |
|
"step": 1076 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.4034, |
|
"step": 1077 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.382, |
|
"step": 1078 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.2569, |
|
"step": 1079 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.2912, |
|
"step": 1080 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 0.0, |
|
"loss": 1.2641, |
|
"step": 1081 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.1642, |
|
"step": 1082 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2204, |
|
"step": 1083 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.3071, |
|
"step": 1084 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2208, |
|
"step": 1085 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2628, |
|
"step": 1086 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.3438, |
|
"step": 1087 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.45, |
|
"step": 1088 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2772, |
|
"step": 1089 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.3997, |
|
"step": 1090 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.411, |
|
"step": 1091 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2209, |
|
"step": 1092 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.1314, |
|
"step": 1093 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2234, |
|
"step": 1094 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2994, |
|
"step": 1095 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2107, |
|
"step": 1096 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 0.0, |
|
"loss": 1.2588, |
|
"step": 1097 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.3409, |
|
"step": 1098 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.3274, |
|
"step": 1099 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.2107, |
|
"step": 1100 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.315, |
|
"step": 1101 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.2599, |
|
"step": 1102 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.2093, |
|
"step": 1103 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.2674, |
|
"step": 1104 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.216, |
|
"step": 1105 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.2827, |
|
"step": 1106 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.1673, |
|
"step": 1107 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.3061, |
|
"step": 1108 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.3474, |
|
"step": 1109 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.2582, |
|
"step": 1110 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.3773, |
|
"step": 1111 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.0822, |
|
"step": 1112 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 0.0, |
|
"loss": 1.2753, |
|
"step": 1113 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2459, |
|
"step": 1114 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.3835, |
|
"step": 1115 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.1639, |
|
"step": 1116 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.333, |
|
"step": 1117 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2792, |
|
"step": 1118 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2931, |
|
"step": 1119 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2091, |
|
"step": 1120 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2477, |
|
"step": 1121 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.1835, |
|
"step": 1122 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.3348, |
|
"step": 1123 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2018, |
|
"step": 1124 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2141, |
|
"step": 1125 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.4496, |
|
"step": 1126 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2375, |
|
"step": 1127 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.3984, |
|
"step": 1128 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.1381, |
|
"step": 1129 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 0.0, |
|
"loss": 1.2685, |
|
"step": 1130 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.3586, |
|
"step": 1131 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.1975, |
|
"step": 1132 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.2319, |
|
"step": 1133 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.2377, |
|
"step": 1134 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.2142, |
|
"step": 1135 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.4997, |
|
"step": 1136 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.1701, |
|
"step": 1137 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.2147, |
|
"step": 1138 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.3302, |
|
"step": 1139 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.2553, |
|
"step": 1140 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.362, |
|
"step": 1141 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.2845, |
|
"step": 1142 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.2671, |
|
"step": 1143 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.1767, |
|
"step": 1144 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.3238, |
|
"step": 1145 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 0.0, |
|
"loss": 1.2355, |
|
"step": 1146 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.1526, |
|
"step": 1147 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.249, |
|
"step": 1148 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.2648, |
|
"step": 1149 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.2615, |
|
"step": 1150 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.3767, |
|
"step": 1151 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.173, |
|
"step": 1152 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.3111, |
|
"step": 1153 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.0724, |
|
"step": 1154 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.3188, |
|
"step": 1155 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.218, |
|
"step": 1156 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.1785, |
|
"step": 1157 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.3291, |
|
"step": 1158 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.2754, |
|
"step": 1159 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.3085, |
|
"step": 1160 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.2005, |
|
"step": 1161 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 0.0, |
|
"loss": 1.5968, |
|
"step": 1162 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.1773, |
|
"step": 1163 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.2263, |
|
"step": 1164 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.2841, |
|
"step": 1165 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.2114, |
|
"step": 1166 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.2448, |
|
"step": 1167 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.5622, |
|
"step": 1168 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.2056, |
|
"step": 1169 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.1521, |
|
"step": 1170 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.4582, |
|
"step": 1171 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.3089, |
|
"step": 1172 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.2573, |
|
"step": 1173 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.3656, |
|
"step": 1174 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.2425, |
|
"step": 1175 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.1986, |
|
"step": 1176 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.3887, |
|
"step": 1177 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 0.0, |
|
"loss": 1.2477, |
|
"step": 1178 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.1427, |
|
"step": 1179 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.2175, |
|
"step": 1180 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.4092, |
|
"step": 1181 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.1418, |
|
"step": 1182 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.3031, |
|
"step": 1183 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.2971, |
|
"step": 1184 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.3914, |
|
"step": 1185 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.3268, |
|
"step": 1186 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.3038, |
|
"step": 1187 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.1588, |
|
"step": 1188 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 0.9393, |
|
"step": 1189 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.3231, |
|
"step": 1190 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.2818, |
|
"step": 1191 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.0893, |
|
"step": 1192 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.3638, |
|
"step": 1193 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.3102, |
|
"step": 1194 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 0.0, |
|
"loss": 1.1931, |
|
"step": 1195 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.1534, |
|
"step": 1196 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.3956, |
|
"step": 1197 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.4381, |
|
"step": 1198 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.4726, |
|
"step": 1199 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.2527, |
|
"step": 1200 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.3108, |
|
"step": 1201 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.3514, |
|
"step": 1202 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.2342, |
|
"step": 1203 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.2232, |
|
"step": 1204 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.2346, |
|
"step": 1205 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.4429, |
|
"step": 1206 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.3612, |
|
"step": 1207 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.2495, |
|
"step": 1208 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.2583, |
|
"step": 1209 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.2281, |
|
"step": 1210 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 0.0, |
|
"loss": 1.1964, |
|
"step": 1211 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.1742, |
|
"step": 1212 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.2384, |
|
"step": 1213 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.2644, |
|
"step": 1214 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.3215, |
|
"step": 1215 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.2653, |
|
"step": 1216 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.1309, |
|
"step": 1217 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.2682, |
|
"step": 1218 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.0355, |
|
"step": 1219 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.3522, |
|
"step": 1220 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.131, |
|
"step": 1221 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.3797, |
|
"step": 1222 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.253, |
|
"step": 1223 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.0941, |
|
"step": 1224 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.3037, |
|
"step": 1225 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.2706, |
|
"step": 1226 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 0.0, |
|
"loss": 1.3062, |
|
"step": 1227 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.21, |
|
"step": 1228 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.322, |
|
"step": 1229 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.2362, |
|
"step": 1230 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.1263, |
|
"step": 1231 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.3367, |
|
"step": 1232 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.2842, |
|
"step": 1233 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.2382, |
|
"step": 1234 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.1986, |
|
"step": 1235 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.3822, |
|
"step": 1236 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.279, |
|
"step": 1237 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.2638, |
|
"step": 1238 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.2648, |
|
"step": 1239 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.1802, |
|
"step": 1240 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.4192, |
|
"step": 1241 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.0685, |
|
"step": 1242 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 0.0, |
|
"loss": 1.4083, |
|
"step": 1243 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.1063, |
|
"step": 1244 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.3261, |
|
"step": 1245 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.3152, |
|
"step": 1246 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.4057, |
|
"step": 1247 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2514, |
|
"step": 1248 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.1074, |
|
"step": 1249 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.1353, |
|
"step": 1250 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.3124, |
|
"step": 1251 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2848, |
|
"step": 1252 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2907, |
|
"step": 1253 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2686, |
|
"step": 1254 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2746, |
|
"step": 1255 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2348, |
|
"step": 1256 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.4894, |
|
"step": 1257 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2858, |
|
"step": 1258 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2884, |
|
"step": 1259 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 0.0, |
|
"loss": 1.2615, |
|
"step": 1260 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3843, |
|
"step": 1261 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3275, |
|
"step": 1262 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3003, |
|
"step": 1263 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.305, |
|
"step": 1264 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.4786, |
|
"step": 1265 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.2223, |
|
"step": 1266 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3913, |
|
"step": 1267 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3453, |
|
"step": 1268 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.2075, |
|
"step": 1269 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3068, |
|
"step": 1270 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3048, |
|
"step": 1271 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.2828, |
|
"step": 1272 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3238, |
|
"step": 1273 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.3292, |
|
"step": 1274 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.2275, |
|
"step": 1275 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 0.0, |
|
"loss": 1.2844, |
|
"step": 1276 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.4134, |
|
"step": 1277 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.1628, |
|
"step": 1278 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.3402, |
|
"step": 1279 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.4187, |
|
"step": 1280 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.1282, |
|
"step": 1281 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.3269, |
|
"step": 1282 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.2887, |
|
"step": 1283 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.2188, |
|
"step": 1284 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.3587, |
|
"step": 1285 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.1813, |
|
"step": 1286 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.2469, |
|
"step": 1287 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.3358, |
|
"step": 1288 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.1821, |
|
"step": 1289 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.1354, |
|
"step": 1290 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.4352, |
|
"step": 1291 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 0.0, |
|
"loss": 1.1648, |
|
"step": 1292 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.328, |
|
"step": 1293 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.4083, |
|
"step": 1294 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.2016, |
|
"step": 1295 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.2978, |
|
"step": 1296 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.3455, |
|
"step": 1297 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.3159, |
|
"step": 1298 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.3621, |
|
"step": 1299 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.3644, |
|
"step": 1300 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.1229, |
|
"step": 1301 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.4087, |
|
"step": 1302 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.3018, |
|
"step": 1303 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.2839, |
|
"step": 1304 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.3037, |
|
"step": 1305 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.2273, |
|
"step": 1306 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.0315, |
|
"step": 1307 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 0.0, |
|
"loss": 1.2416, |
|
"step": 1308 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.3961, |
|
"step": 1309 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.2927, |
|
"step": 1310 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.4131, |
|
"step": 1311 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.1581, |
|
"step": 1312 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.2914, |
|
"step": 1313 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.4193, |
|
"step": 1314 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.3016, |
|
"step": 1315 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.3346, |
|
"step": 1316 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.3861, |
|
"step": 1317 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.2041, |
|
"step": 1318 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.2784, |
|
"step": 1319 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.1276, |
|
"step": 1320 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.121, |
|
"step": 1321 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.2583, |
|
"step": 1322 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.4013, |
|
"step": 1323 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.4751, |
|
"step": 1324 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 0.0, |
|
"loss": 1.3687, |
|
"step": 1325 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.343, |
|
"step": 1326 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.2615, |
|
"step": 1327 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.3691, |
|
"step": 1328 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.259, |
|
"step": 1329 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.171, |
|
"step": 1330 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.1405, |
|
"step": 1331 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.3643, |
|
"step": 1332 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.2387, |
|
"step": 1333 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.3776, |
|
"step": 1334 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.3048, |
|
"step": 1335 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.3251, |
|
"step": 1336 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.5075, |
|
"step": 1337 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.4189, |
|
"step": 1338 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.2672, |
|
"step": 1339 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.124, |
|
"step": 1340 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 0.0, |
|
"loss": 1.446, |
|
"step": 1341 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.2823, |
|
"step": 1342 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.3723, |
|
"step": 1343 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.1703, |
|
"step": 1344 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.1446, |
|
"step": 1345 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.2875, |
|
"step": 1346 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.4609, |
|
"step": 1347 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.1763, |
|
"step": 1348 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.2255, |
|
"step": 1349 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.3169, |
|
"step": 1350 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.1857, |
|
"step": 1351 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.4281, |
|
"step": 1352 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.1975, |
|
"step": 1353 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.2722, |
|
"step": 1354 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.4018, |
|
"step": 1355 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.2439, |
|
"step": 1356 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 0.0, |
|
"loss": 1.3282, |
|
"step": 1357 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.509, |
|
"step": 1358 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.3154, |
|
"step": 1359 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.4223, |
|
"step": 1360 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.2517, |
|
"step": 1361 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.2955, |
|
"step": 1362 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.3627, |
|
"step": 1363 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.5124, |
|
"step": 1364 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.3483, |
|
"step": 1365 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.3437, |
|
"step": 1366 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.2165, |
|
"step": 1367 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.2491, |
|
"step": 1368 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.2627, |
|
"step": 1369 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.2343, |
|
"step": 1370 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.2444, |
|
"step": 1371 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.3318, |
|
"step": 1372 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 0.0, |
|
"loss": 1.4575, |
|
"step": 1373 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.1414, |
|
"step": 1374 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.3893, |
|
"step": 1375 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.2013, |
|
"step": 1376 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.182, |
|
"step": 1377 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.3595, |
|
"step": 1378 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.2267, |
|
"step": 1379 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.2235, |
|
"step": 1380 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.3763, |
|
"step": 1381 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.1283, |
|
"step": 1382 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.2477, |
|
"step": 1383 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.286, |
|
"step": 1384 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.3129, |
|
"step": 1385 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.2616, |
|
"step": 1386 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.2608, |
|
"step": 1387 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.3496, |
|
"step": 1388 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.2916, |
|
"step": 1389 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 0.0, |
|
"loss": 1.2275, |
|
"step": 1390 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.3341, |
|
"step": 1391 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.2646, |
|
"step": 1392 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.3133, |
|
"step": 1393 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.3635, |
|
"step": 1394 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.279, |
|
"step": 1395 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.4758, |
|
"step": 1396 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.4896, |
|
"step": 1397 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.2851, |
|
"step": 1398 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.2935, |
|
"step": 1399 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.3487, |
|
"step": 1400 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.2814, |
|
"step": 1401 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.4303, |
|
"step": 1402 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.3196, |
|
"step": 1403 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.3638, |
|
"step": 1404 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.4937, |
|
"step": 1405 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 0.0, |
|
"loss": 1.4708, |
|
"step": 1406 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.236, |
|
"step": 1407 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.4315, |
|
"step": 1408 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.2594, |
|
"step": 1409 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.3253, |
|
"step": 1410 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.4172, |
|
"step": 1411 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.3106, |
|
"step": 1412 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.2676, |
|
"step": 1413 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.2459, |
|
"step": 1414 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.3074, |
|
"step": 1415 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.5803, |
|
"step": 1416 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.4103, |
|
"step": 1417 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.1719, |
|
"step": 1418 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.3392, |
|
"step": 1419 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.2417, |
|
"step": 1420 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.3102, |
|
"step": 1421 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 0.0, |
|
"loss": 1.239, |
|
"step": 1422 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.4267, |
|
"step": 1423 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.3567, |
|
"step": 1424 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.1479, |
|
"step": 1425 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.4659, |
|
"step": 1426 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.347, |
|
"step": 1427 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.3434, |
|
"step": 1428 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.2177, |
|
"step": 1429 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.3709, |
|
"step": 1430 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.448, |
|
"step": 1431 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.2925, |
|
"step": 1432 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.1606, |
|
"step": 1433 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.163, |
|
"step": 1434 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.4178, |
|
"step": 1435 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.5311, |
|
"step": 1436 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.2456, |
|
"step": 1437 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.2234, |
|
"step": 1438 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 0.0, |
|
"loss": 1.2701, |
|
"step": 1439 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.2305, |
|
"step": 1440 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.2948, |
|
"step": 1441 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.2812, |
|
"step": 1442 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.4094, |
|
"step": 1443 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.4507, |
|
"step": 1444 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.2405, |
|
"step": 1445 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.3447, |
|
"step": 1446 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.3212, |
|
"step": 1447 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.385, |
|
"step": 1448 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.3304, |
|
"step": 1449 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.3339, |
|
"step": 1450 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.4386, |
|
"step": 1451 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.4017, |
|
"step": 1452 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.355, |
|
"step": 1453 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.1572, |
|
"step": 1454 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 0.0, |
|
"loss": 1.4216, |
|
"step": 1455 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.38, |
|
"step": 1456 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.3569, |
|
"step": 1457 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.2598, |
|
"step": 1458 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.4702, |
|
"step": 1459 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.1405, |
|
"step": 1460 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.3461, |
|
"step": 1461 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.3709, |
|
"step": 1462 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.2258, |
|
"step": 1463 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.2995, |
|
"step": 1464 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.3896, |
|
"step": 1465 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.3072, |
|
"step": 1466 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.4293, |
|
"step": 1467 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.2015, |
|
"step": 1468 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.4695, |
|
"step": 1469 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.1791, |
|
"step": 1470 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 0.0, |
|
"loss": 1.4435, |
|
"step": 1471 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.3438, |
|
"step": 1472 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.2363, |
|
"step": 1473 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.382, |
|
"step": 1474 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.2145, |
|
"step": 1475 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.3642, |
|
"step": 1476 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.2463, |
|
"step": 1477 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.355, |
|
"step": 1478 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.2366, |
|
"step": 1479 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.3401, |
|
"step": 1480 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.3895, |
|
"step": 1481 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.1932, |
|
"step": 1482 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.3237, |
|
"step": 1483 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.2219, |
|
"step": 1484 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.3928, |
|
"step": 1485 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.3802, |
|
"step": 1486 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 0.0, |
|
"loss": 1.2238, |
|
"step": 1487 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.305, |
|
"step": 1488 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.3448, |
|
"step": 1489 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.2038, |
|
"step": 1490 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.2917, |
|
"step": 1491 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.2026, |
|
"step": 1492 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.2306, |
|
"step": 1493 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.2146, |
|
"step": 1494 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.3176, |
|
"step": 1495 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.5465, |
|
"step": 1496 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.2335, |
|
"step": 1497 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.2875, |
|
"step": 1498 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.1973, |
|
"step": 1499 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.4062, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.374, |
|
"step": 1501 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.3856, |
|
"step": 1502 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.3272, |
|
"step": 1503 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 0.0, |
|
"loss": 1.1662, |
|
"step": 1504 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.2478, |
|
"step": 1505 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.2458, |
|
"step": 1506 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.2229, |
|
"step": 1507 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.1854, |
|
"step": 1508 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.3922, |
|
"step": 1509 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.2329, |
|
"step": 1510 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.3555, |
|
"step": 1511 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.3927, |
|
"step": 1512 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.2795, |
|
"step": 1513 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.467, |
|
"step": 1514 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.4493, |
|
"step": 1515 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.3017, |
|
"step": 1516 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.3762, |
|
"step": 1517 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.1526, |
|
"step": 1518 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.3726, |
|
"step": 1519 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 0.0, |
|
"loss": 1.5151, |
|
"step": 1520 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.2724, |
|
"step": 1521 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.37, |
|
"step": 1522 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.1539, |
|
"step": 1523 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.3598, |
|
"step": 1524 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.3767, |
|
"step": 1525 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.2058, |
|
"step": 1526 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.4847, |
|
"step": 1527 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.2806, |
|
"step": 1528 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.3664, |
|
"step": 1529 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.4201, |
|
"step": 1530 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.444, |
|
"step": 1531 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.3904, |
|
"step": 1532 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.2657, |
|
"step": 1533 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.4583, |
|
"step": 1534 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.3211, |
|
"step": 1535 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 0.0, |
|
"loss": 1.3342, |
|
"step": 1536 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.2621, |
|
"step": 1537 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.3172, |
|
"step": 1538 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.3282, |
|
"step": 1539 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.3931, |
|
"step": 1540 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.2736, |
|
"step": 1541 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.417, |
|
"step": 1542 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.2688, |
|
"step": 1543 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.2682, |
|
"step": 1544 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.3287, |
|
"step": 1545 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.3805, |
|
"step": 1546 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.4528, |
|
"step": 1547 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.3783, |
|
"step": 1548 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.1474, |
|
"step": 1549 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.2024, |
|
"step": 1550 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.2026, |
|
"step": 1551 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 0.0, |
|
"loss": 1.3891, |
|
"step": 1552 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.3637, |
|
"step": 1553 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.4384, |
|
"step": 1554 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.3686, |
|
"step": 1555 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.4883, |
|
"step": 1556 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.4316, |
|
"step": 1557 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.2798, |
|
"step": 1558 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.2609, |
|
"step": 1559 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.2243, |
|
"step": 1560 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.2158, |
|
"step": 1561 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.3825, |
|
"step": 1562 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.4618, |
|
"step": 1563 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.2915, |
|
"step": 1564 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.4045, |
|
"step": 1565 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.2909, |
|
"step": 1566 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.1258, |
|
"step": 1567 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.4053, |
|
"step": 1568 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 0.0, |
|
"loss": 1.3747, |
|
"step": 1569 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.2539, |
|
"step": 1570 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.2615, |
|
"step": 1571 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.4086, |
|
"step": 1572 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.3772, |
|
"step": 1573 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.4835, |
|
"step": 1574 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.1869, |
|
"step": 1575 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.3675, |
|
"step": 1576 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.4689, |
|
"step": 1577 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.4215, |
|
"step": 1578 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.2765, |
|
"step": 1579 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.3309, |
|
"step": 1580 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.3971, |
|
"step": 1581 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.42, |
|
"step": 1582 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.1888, |
|
"step": 1583 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.3269, |
|
"step": 1584 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 0.0, |
|
"loss": 1.2317, |
|
"step": 1585 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.3863, |
|
"step": 1586 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.4418, |
|
"step": 1587 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.3152, |
|
"step": 1588 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.2346, |
|
"step": 1589 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.3975, |
|
"step": 1590 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.359, |
|
"step": 1591 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.4831, |
|
"step": 1592 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.1941, |
|
"step": 1593 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.4048, |
|
"step": 1594 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.2892, |
|
"step": 1595 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.1816, |
|
"step": 1596 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.362, |
|
"step": 1597 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.3701, |
|
"step": 1598 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.3358, |
|
"step": 1599 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.2318, |
|
"step": 1600 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 0.0, |
|
"loss": 1.0476, |
|
"step": 1601 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.4123, |
|
"step": 1602 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.3312, |
|
"step": 1603 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.3927, |
|
"step": 1604 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.4173, |
|
"step": 1605 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.3974, |
|
"step": 1606 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.1858, |
|
"step": 1607 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.4656, |
|
"step": 1608 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.3686, |
|
"step": 1609 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.4562, |
|
"step": 1610 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.4116, |
|
"step": 1611 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.3388, |
|
"step": 1612 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.4117, |
|
"step": 1613 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.3075, |
|
"step": 1614 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.3134, |
|
"step": 1615 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.5144, |
|
"step": 1616 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 0.0, |
|
"loss": 1.4333, |
|
"step": 1617 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.4132, |
|
"step": 1618 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.2455, |
|
"step": 1619 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.2306, |
|
"step": 1620 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.2623, |
|
"step": 1621 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.4016, |
|
"step": 1622 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.4198, |
|
"step": 1623 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 0.0, |
|
"loss": 2.0855, |
|
"step": 1624 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 0.0, |
|
"loss": 1.2834, |
|
"step": 1625 |
|
} |
|
], |
|
"logging_steps": 1.0, |
|
"max_steps": 6000, |
|
"num_train_epochs": 4, |
|
"save_steps": 1625, |
|
"total_flos": 5.114647624548352e+16, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|