| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 2.4627336358741636, | |
| "global_step": 1000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.0000000000000002e-07, | |
| "loss": 2.0749, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4.0000000000000003e-07, | |
| "loss": 2.0939, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.000000000000001e-07, | |
| "loss": 2.0808, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.000000000000001e-07, | |
| "loss": 2.0523, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.0000000000000002e-06, | |
| "loss": 2.0731, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.2000000000000002e-06, | |
| "loss": 2.0781, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 2.0222, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.6000000000000001e-06, | |
| "loss": 2.0379, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.8000000000000001e-06, | |
| "loss": 2.0547, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.0000000000000003e-06, | |
| "loss": 2.088, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.2e-06, | |
| "loss": 2.0938, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 1.9722, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.6e-06, | |
| "loss": 2.0697, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.8000000000000003e-06, | |
| "loss": 2.0278, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3e-06, | |
| "loss": 2.0449, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.2000000000000003e-06, | |
| "loss": 2.0585, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 2.0547, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.6000000000000003e-06, | |
| "loss": 2.0735, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 3.8000000000000005e-06, | |
| "loss": 2.1028, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 2.1025, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.2000000000000004e-06, | |
| "loss": 2.0129, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.4e-06, | |
| "loss": 2.0452, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.600000000000001e-06, | |
| "loss": 2.0978, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.800000000000001e-06, | |
| "loss": 2.0609, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5e-06, | |
| "loss": 2.0783, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.2e-06, | |
| "loss": 2.0535, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 2.0383, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.600000000000001e-06, | |
| "loss": 2.0492, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.8e-06, | |
| "loss": 2.097, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6e-06, | |
| "loss": 2.0641, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 6.200000000000001e-06, | |
| "loss": 2.0511, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 2.0927, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 6.600000000000001e-06, | |
| "loss": 1.9922, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 6.800000000000001e-06, | |
| "loss": 2.0557, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 7e-06, | |
| "loss": 1.992, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 7.2000000000000005e-06, | |
| "loss": 2.0997, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 7.4e-06, | |
| "loss": 2.0116, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 7.600000000000001e-06, | |
| "loss": 1.9953, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 7.800000000000002e-06, | |
| "loss": 2.043, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 2.0136, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 8.2e-06, | |
| "loss": 2.0295, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 2.0217, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 8.6e-06, | |
| "loss": 2.0433, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 8.8e-06, | |
| "loss": 1.937, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9e-06, | |
| "loss": 1.9604, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.200000000000002e-06, | |
| "loss": 1.9755, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.4e-06, | |
| "loss": 1.9808, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.600000000000001e-06, | |
| "loss": 2.0126, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.800000000000001e-06, | |
| "loss": 2.0072, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1e-05, | |
| "loss": 2.0112, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.02e-05, | |
| "loss": 2.0066, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.04e-05, | |
| "loss": 1.9487, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.0600000000000002e-05, | |
| "loss": 2.0321, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.0800000000000002e-05, | |
| "loss": 1.9578, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.1000000000000001e-05, | |
| "loss": 2.0002, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.1200000000000001e-05, | |
| "loss": 1.9924, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.14e-05, | |
| "loss": 1.9803, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.16e-05, | |
| "loss": 1.9154, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.18e-05, | |
| "loss": 1.9307, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.2e-05, | |
| "loss": 1.9131, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.22e-05, | |
| "loss": 1.9555, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.2400000000000002e-05, | |
| "loss": 1.9246, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.2600000000000001e-05, | |
| "loss": 1.9316, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.2800000000000001e-05, | |
| "loss": 1.8861, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.3000000000000001e-05, | |
| "loss": 1.9391, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.3200000000000002e-05, | |
| "loss": 1.8988, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.3400000000000002e-05, | |
| "loss": 1.8759, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.3600000000000002e-05, | |
| "loss": 1.8406, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.38e-05, | |
| "loss": 1.8932, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.4e-05, | |
| "loss": 1.8809, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.4200000000000001e-05, | |
| "loss": 1.877, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.4400000000000001e-05, | |
| "loss": 1.8344, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.46e-05, | |
| "loss": 1.8449, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.48e-05, | |
| "loss": 1.7725, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.5000000000000002e-05, | |
| "loss": 1.7643, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.5200000000000002e-05, | |
| "loss": 1.7966, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.54e-05, | |
| "loss": 1.8069, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.5600000000000003e-05, | |
| "loss": 1.7931, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.58e-05, | |
| "loss": 1.7639, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.6000000000000003e-05, | |
| "loss": 1.7514, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.62e-05, | |
| "loss": 1.7457, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.64e-05, | |
| "loss": 1.7195, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.66e-05, | |
| "loss": 1.7307, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.6800000000000002e-05, | |
| "loss": 1.6572, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.7e-05, | |
| "loss": 1.6728, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.72e-05, | |
| "loss": 1.6499, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.7400000000000003e-05, | |
| "loss": 1.6669, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.76e-05, | |
| "loss": 1.5629, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.7800000000000002e-05, | |
| "loss": 1.5887, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8e-05, | |
| "loss": 1.5662, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8200000000000002e-05, | |
| "loss": 1.574, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8400000000000003e-05, | |
| "loss": 1.5435, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.86e-05, | |
| "loss": 1.5675, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.88e-05, | |
| "loss": 1.5504, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.9e-05, | |
| "loss": 1.5344, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.9200000000000003e-05, | |
| "loss": 1.5012, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.94e-05, | |
| "loss": 1.4572, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.9600000000000002e-05, | |
| "loss": 1.4746, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.98e-05, | |
| "loss": 1.3997, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2e-05, | |
| "loss": 1.4291, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.9982110912343474e-05, | |
| "loss": 1.3956, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.9964221824686943e-05, | |
| "loss": 1.3893, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.9946332737030415e-05, | |
| "loss": 1.3612, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9928443649373884e-05, | |
| "loss": 1.3851, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9910554561717356e-05, | |
| "loss": 1.4022, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9892665474060825e-05, | |
| "loss": 1.3556, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9874776386404297e-05, | |
| "loss": 1.3112, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9856887298747766e-05, | |
| "loss": 1.3367, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9838998211091234e-05, | |
| "loss": 1.2876, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9821109123434707e-05, | |
| "loss": 1.2824, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9803220035778175e-05, | |
| "loss": 1.2417, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9785330948121648e-05, | |
| "loss": 1.2683, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9767441860465116e-05, | |
| "loss": 1.2332, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.974955277280859e-05, | |
| "loss": 1.2492, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9731663685152057e-05, | |
| "loss": 1.2397, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.971377459749553e-05, | |
| "loss": 1.2537, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.9695885509839e-05, | |
| "loss": 1.2242, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.967799642218247e-05, | |
| "loss": 1.211, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.9660107334525943e-05, | |
| "loss": 1.1657, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.964221824686941e-05, | |
| "loss": 1.1705, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9624329159212884e-05, | |
| "loss": 1.1687, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9606440071556353e-05, | |
| "loss": 1.1469, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9588550983899825e-05, | |
| "loss": 1.1621, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.9570661896243294e-05, | |
| "loss": 1.1559, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.9552772808586762e-05, | |
| "loss": 1.1086, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.9534883720930235e-05, | |
| "loss": 1.089, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.9516994633273703e-05, | |
| "loss": 1.0888, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9499105545617176e-05, | |
| "loss": 1.0849, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9481216457960644e-05, | |
| "loss": 1.0562, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9463327370304117e-05, | |
| "loss": 1.0523, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9445438282647585e-05, | |
| "loss": 1.0164, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9427549194991058e-05, | |
| "loss": 1.0467, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9409660107334526e-05, | |
| "loss": 1.0242, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9391771019678e-05, | |
| "loss": 1.0218, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9373881932021467e-05, | |
| "loss": 0.988, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.935599284436494e-05, | |
| "loss": 1.0138, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.933810375670841e-05, | |
| "loss": 1.0288, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.932021466905188e-05, | |
| "loss": 0.9849, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9302325581395353e-05, | |
| "loss": 1.0165, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.928443649373882e-05, | |
| "loss": 0.979, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9266547406082294e-05, | |
| "loss": 0.9418, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9248658318425763e-05, | |
| "loss": 0.9832, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.923076923076923e-05, | |
| "loss": 0.9422, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9212880143112704e-05, | |
| "loss": 1.0011, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9194991055456172e-05, | |
| "loss": 0.9562, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.917710196779964e-05, | |
| "loss": 0.9214, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9159212880143113e-05, | |
| "loss": 0.9713, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9141323792486586e-05, | |
| "loss": 0.9227, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9123434704830054e-05, | |
| "loss": 0.944, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9105545617173527e-05, | |
| "loss": 0.9623, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9087656529516995e-05, | |
| "loss": 0.9324, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9069767441860468e-05, | |
| "loss": 0.941, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9051878354203936e-05, | |
| "loss": 0.9442, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.903398926654741e-05, | |
| "loss": 0.8992, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9016100178890877e-05, | |
| "loss": 0.8882, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.899821109123435e-05, | |
| "loss": 0.9508, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.8980322003577822e-05, | |
| "loss": 0.9295, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.896243291592129e-05, | |
| "loss": 0.9519, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.894454382826476e-05, | |
| "loss": 0.9342, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.892665474060823e-05, | |
| "loss": 0.9766, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.89087656529517e-05, | |
| "loss": 0.9625, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.889087656529517e-05, | |
| "loss": 0.8774, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.887298747763864e-05, | |
| "loss": 0.9137, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.885509838998211e-05, | |
| "loss": 0.9574, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.8837209302325582e-05, | |
| "loss": 0.9246, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.881932021466905e-05, | |
| "loss": 0.8908, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.8801431127012523e-05, | |
| "loss": 0.8991, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.8783542039355996e-05, | |
| "loss": 0.9756, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.8765652951699464e-05, | |
| "loss": 0.9422, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.8747763864042937e-05, | |
| "loss": 0.9082, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.8729874776386405e-05, | |
| "loss": 0.9071, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.8711985688729878e-05, | |
| "loss": 0.9027, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.8694096601073346e-05, | |
| "loss": 0.8683, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.867620751341682e-05, | |
| "loss": 0.8829, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.8658318425760287e-05, | |
| "loss": 0.9089, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.864042933810376e-05, | |
| "loss": 0.9022, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.862254025044723e-05, | |
| "loss": 0.8839, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.86046511627907e-05, | |
| "loss": 0.9115, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.858676207513417e-05, | |
| "loss": 0.869, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.8568872987477638e-05, | |
| "loss": 0.9228, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.855098389982111e-05, | |
| "loss": 0.9513, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.853309481216458e-05, | |
| "loss": 0.8936, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.851520572450805e-05, | |
| "loss": 0.883, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.849731663685152e-05, | |
| "loss": 0.8078, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.8479427549194992e-05, | |
| "loss": 0.8923, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.8461538461538465e-05, | |
| "loss": 0.8929, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.8443649373881933e-05, | |
| "loss": 0.889, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.8425760286225406e-05, | |
| "loss": 0.8927, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.8407871198568874e-05, | |
| "loss": 0.8774, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.8389982110912347e-05, | |
| "loss": 0.8691, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.8372093023255815e-05, | |
| "loss": 0.9064, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.8354203935599288e-05, | |
| "loss": 0.8597, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.8336314847942756e-05, | |
| "loss": 0.8927, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.831842576028623e-05, | |
| "loss": 0.8337, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.8300536672629697e-05, | |
| "loss": 0.8601, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.8282647584973166e-05, | |
| "loss": 0.8689, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.826475849731664e-05, | |
| "loss": 0.8508, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.8246869409660107e-05, | |
| "loss": 0.8423, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.822898032200358e-05, | |
| "loss": 0.8897, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.8211091234347048e-05, | |
| "loss": 0.867, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.819320214669052e-05, | |
| "loss": 0.8933, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.817531305903399e-05, | |
| "loss": 0.9027, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.815742397137746e-05, | |
| "loss": 0.8983, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.813953488372093e-05, | |
| "loss": 0.8857, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.8121645796064402e-05, | |
| "loss": 0.8266, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.8103756708407875e-05, | |
| "loss": 0.8404, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.8085867620751343e-05, | |
| "loss": 0.8731, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.8067978533094816e-05, | |
| "loss": 0.8429, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.8050089445438284e-05, | |
| "loss": 0.8586, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.8032200357781757e-05, | |
| "loss": 0.8492, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.8014311270125225e-05, | |
| "loss": 0.8351, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.7996422182468694e-05, | |
| "loss": 0.8566, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.7978533094812166e-05, | |
| "loss": 0.872, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.7960644007155635e-05, | |
| "loss": 0.8946, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.7942754919499107e-05, | |
| "loss": 0.8784, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.7924865831842576e-05, | |
| "loss": 0.8371, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.790697674418605e-05, | |
| "loss": 0.8649, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.7889087656529517e-05, | |
| "loss": 0.819, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.787119856887299e-05, | |
| "loss": 0.8903, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.7853309481216458e-05, | |
| "loss": 0.8675, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.783542039355993e-05, | |
| "loss": 0.8728, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.78175313059034e-05, | |
| "loss": 0.8753, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.779964221824687e-05, | |
| "loss": 0.8156, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.7781753130590344e-05, | |
| "loss": 0.8388, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.7763864042933812e-05, | |
| "loss": 0.8928, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.7745974955277285e-05, | |
| "loss": 0.8225, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.7728085867620753e-05, | |
| "loss": 0.8366, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.7710196779964222e-05, | |
| "loss": 0.8131, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.7692307692307694e-05, | |
| "loss": 0.8717, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.7674418604651163e-05, | |
| "loss": 0.8281, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.7656529516994635e-05, | |
| "loss": 0.8203, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.7638640429338104e-05, | |
| "loss": 0.8409, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.7620751341681576e-05, | |
| "loss": 0.9037, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.7602862254025045e-05, | |
| "loss": 0.9087, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.7584973166368517e-05, | |
| "loss": 0.8616, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.7567084078711986e-05, | |
| "loss": 0.8293, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.754919499105546e-05, | |
| "loss": 0.9028, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.7531305903398927e-05, | |
| "loss": 0.8728, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.75134168157424e-05, | |
| "loss": 0.8617, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.7495527728085868e-05, | |
| "loss": 0.8362, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.747763864042934e-05, | |
| "loss": 0.9042, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.745974955277281e-05, | |
| "loss": 0.8496, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.744186046511628e-05, | |
| "loss": 0.8953, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.742397137745975e-05, | |
| "loss": 0.8925, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.7406082289803222e-05, | |
| "loss": 0.8732, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.738819320214669e-05, | |
| "loss": 0.8479, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.7370304114490163e-05, | |
| "loss": 0.841, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.7352415026833632e-05, | |
| "loss": 0.8621, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.73345259391771e-05, | |
| "loss": 0.8883, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.7316636851520573e-05, | |
| "loss": 0.9141, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.7298747763864042e-05, | |
| "loss": 0.8723, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.7280858676207514e-05, | |
| "loss": 0.8421, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.7262969588550986e-05, | |
| "loss": 0.81, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.7245080500894455e-05, | |
| "loss": 0.8013, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.7227191413237927e-05, | |
| "loss": 0.8855, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.7209302325581396e-05, | |
| "loss": 0.8604, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.719141323792487e-05, | |
| "loss": 0.8795, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.7173524150268337e-05, | |
| "loss": 0.8324, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.715563506261181e-05, | |
| "loss": 0.8615, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.7137745974955278e-05, | |
| "loss": 0.8045, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.711985688729875e-05, | |
| "loss": 0.8664, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.710196779964222e-05, | |
| "loss": 0.848, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.708407871198569e-05, | |
| "loss": 0.8489, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.706618962432916e-05, | |
| "loss": 0.8425, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.704830053667263e-05, | |
| "loss": 0.881, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.70304114490161e-05, | |
| "loss": 0.877, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.701252236135957e-05, | |
| "loss": 0.8441, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.6994633273703042e-05, | |
| "loss": 0.8549, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.697674418604651e-05, | |
| "loss": 0.8613, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.6958855098389983e-05, | |
| "loss": 0.8643, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.6940966010733455e-05, | |
| "loss": 0.87, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.6923076923076924e-05, | |
| "loss": 0.9223, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.6905187835420396e-05, | |
| "loss": 0.8748, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.6887298747763865e-05, | |
| "loss": 0.8234, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.6869409660107337e-05, | |
| "loss": 0.8259, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.6851520572450806e-05, | |
| "loss": 0.8406, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.683363148479428e-05, | |
| "loss": 0.8218, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.6815742397137747e-05, | |
| "loss": 0.8564, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.679785330948122e-05, | |
| "loss": 0.885, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.6779964221824688e-05, | |
| "loss": 0.8218, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.6762075134168157e-05, | |
| "loss": 0.8206, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.674418604651163e-05, | |
| "loss": 0.8428, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.6726296958855098e-05, | |
| "loss": 0.8816, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.670840787119857e-05, | |
| "loss": 0.8458, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.669051878354204e-05, | |
| "loss": 0.8412, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.667262969588551e-05, | |
| "loss": 0.8458, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.665474060822898e-05, | |
| "loss": 0.854, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.6636851520572452e-05, | |
| "loss": 0.8831, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.661896243291592e-05, | |
| "loss": 0.9012, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.6601073345259393e-05, | |
| "loss": 0.8148, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.6583184257602865e-05, | |
| "loss": 0.8257, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.6565295169946334e-05, | |
| "loss": 0.8451, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.6547406082289806e-05, | |
| "loss": 0.8355, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.6529516994633275e-05, | |
| "loss": 0.8564, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.6511627906976747e-05, | |
| "loss": 0.8922, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.6493738819320216e-05, | |
| "loss": 0.87, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.647584973166369e-05, | |
| "loss": 0.8307, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.6457960644007157e-05, | |
| "loss": 0.8233, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.6440071556350626e-05, | |
| "loss": 0.8436, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.6422182468694098e-05, | |
| "loss": 0.8316, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.6404293381037567e-05, | |
| "loss": 0.8468, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.638640429338104e-05, | |
| "loss": 0.851, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6368515205724508e-05, | |
| "loss": 0.8277, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.635062611806798e-05, | |
| "loss": 0.8296, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.633273703041145e-05, | |
| "loss": 0.8571, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.631484794275492e-05, | |
| "loss": 0.8691, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.629695885509839e-05, | |
| "loss": 0.8477, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6279069767441862e-05, | |
| "loss": 0.8399, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6261180679785334e-05, | |
| "loss": 0.8243, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6243291592128803e-05, | |
| "loss": 0.8439, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6225402504472275e-05, | |
| "loss": 0.8175, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6207513416815744e-05, | |
| "loss": 0.8134, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6189624329159216e-05, | |
| "loss": 0.8575, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6171735241502685e-05, | |
| "loss": 0.8559, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6153846153846154e-05, | |
| "loss": 0.8374, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6135957066189626e-05, | |
| "loss": 0.7871, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6118067978533095e-05, | |
| "loss": 0.8712, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6100178890876567e-05, | |
| "loss": 0.8224, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6082289803220036e-05, | |
| "loss": 0.8406, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6064400715563508e-05, | |
| "loss": 0.8155, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6046511627906977e-05, | |
| "loss": 0.8617, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.602862254025045e-05, | |
| "loss": 0.8328, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.6010733452593918e-05, | |
| "loss": 0.829, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.599284436493739e-05, | |
| "loss": 0.8293, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.597495527728086e-05, | |
| "loss": 0.8227, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.595706618962433e-05, | |
| "loss": 0.8653, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.59391771019678e-05, | |
| "loss": 0.8151, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5921288014311272e-05, | |
| "loss": 0.7727, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5903398926654744e-05, | |
| "loss": 0.862, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5885509838998213e-05, | |
| "loss": 0.7711, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5867620751341682e-05, | |
| "loss": 0.7793, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5849731663685154e-05, | |
| "loss": 0.8516, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5831842576028623e-05, | |
| "loss": 0.8307, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5813953488372095e-05, | |
| "loss": 0.8396, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5796064400715564e-05, | |
| "loss": 0.8286, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5778175313059033e-05, | |
| "loss": 0.8009, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5760286225402505e-05, | |
| "loss": 0.8019, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5742397137745977e-05, | |
| "loss": 0.8213, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5724508050089446e-05, | |
| "loss": 0.8219, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5706618962432918e-05, | |
| "loss": 0.8349, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5688729874776387e-05, | |
| "loss": 0.8143, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.567084078711986e-05, | |
| "loss": 0.8195, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5652951699463328e-05, | |
| "loss": 0.7934, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.56350626118068e-05, | |
| "loss": 0.8411, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.561717352415027e-05, | |
| "loss": 0.8399, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.559928443649374e-05, | |
| "loss": 0.8198, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.558139534883721e-05, | |
| "loss": 0.8247, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.5563506261180682e-05, | |
| "loss": 0.8273, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.554561717352415e-05, | |
| "loss": 0.8559, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.5527728085867623e-05, | |
| "loss": 0.8711, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.5509838998211092e-05, | |
| "loss": 0.7908, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.549194991055456e-05, | |
| "loss": 0.77, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.5474060822898033e-05, | |
| "loss": 0.8534, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.5456171735241502e-05, | |
| "loss": 0.8307, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.5438282647584974e-05, | |
| "loss": 0.833, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5420393559928446e-05, | |
| "loss": 0.8106, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5402504472271915e-05, | |
| "loss": 0.8492, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5384615384615387e-05, | |
| "loss": 0.8324, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5366726296958856e-05, | |
| "loss": 0.7787, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.5348837209302328e-05, | |
| "loss": 0.8056, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.5330948121645797e-05, | |
| "loss": 0.8535, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.531305903398927e-05, | |
| "loss": 0.8351, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.5295169946332738e-05, | |
| "loss": 0.8457, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.527728085867621e-05, | |
| "loss": 0.8336, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.525939177101968e-05, | |
| "loss": 0.8459, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.524150268336315e-05, | |
| "loss": 0.8714, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.5223613595706622e-05, | |
| "loss": 0.8258, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.520572450805009e-05, | |
| "loss": 0.9213, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.5187835420393563e-05, | |
| "loss": 0.8341, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.5169946332737031e-05, | |
| "loss": 0.8022, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.5152057245080502e-05, | |
| "loss": 0.7943, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.5134168157423972e-05, | |
| "loss": 0.852, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.5116279069767443e-05, | |
| "loss": 0.8684, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.5098389982110912e-05, | |
| "loss": 0.8951, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.5080500894454384e-05, | |
| "loss": 0.8065, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.5062611806797856e-05, | |
| "loss": 0.8308, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.5044722719141325e-05, | |
| "loss": 0.8609, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.5026833631484795e-05, | |
| "loss": 0.8266, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.5008944543828266e-05, | |
| "loss": 0.8265, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.4991055456171736e-05, | |
| "loss": 0.8416, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.4973166368515207e-05, | |
| "loss": 0.8167, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.4955277280858677e-05, | |
| "loss": 0.8447, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.4937388193202146e-05, | |
| "loss": 0.8328, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4919499105545618e-05, | |
| "loss": 0.8347, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.490161001788909e-05, | |
| "loss": 0.8552, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.488372093023256e-05, | |
| "loss": 0.8543, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.486583184257603e-05, | |
| "loss": 0.8652, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.48479427549195e-05, | |
| "loss": 0.8058, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.4830053667262971e-05, | |
| "loss": 0.8092, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.481216457960644e-05, | |
| "loss": 0.8588, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.4794275491949912e-05, | |
| "loss": 0.8591, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.477638640429338e-05, | |
| "loss": 0.8499, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.4758497316636853e-05, | |
| "loss": 0.8223, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.4740608228980323e-05, | |
| "loss": 0.8023, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.4722719141323794e-05, | |
| "loss": 0.8077, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.4704830053667264e-05, | |
| "loss": 0.8578, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.4686940966010735e-05, | |
| "loss": 0.8287, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.4669051878354205e-05, | |
| "loss": 0.8481, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.4651162790697674e-05, | |
| "loss": 0.8197, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.4633273703041146e-05, | |
| "loss": 0.811, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.4615384615384615e-05, | |
| "loss": 0.7848, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.4597495527728087e-05, | |
| "loss": 0.7713, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.4579606440071556e-05, | |
| "loss": 0.8476, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.4561717352415028e-05, | |
| "loss": 0.829, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.4543828264758499e-05, | |
| "loss": 0.8703, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.452593917710197e-05, | |
| "loss": 0.8499, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.450805008944544e-05, | |
| "loss": 1.0286, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.4490161001788909e-05, | |
| "loss": 0.817, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.4472271914132381e-05, | |
| "loss": 0.8247, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.445438282647585e-05, | |
| "loss": 0.8081, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.4436493738819322e-05, | |
| "loss": 0.8309, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.441860465116279e-05, | |
| "loss": 0.8615, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.4400715563506263e-05, | |
| "loss": 0.8196, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.4382826475849733e-05, | |
| "loss": 0.8586, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.4364937388193202e-05, | |
| "loss": 0.8293, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.4347048300536674e-05, | |
| "loss": 0.8287, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.4329159212880143e-05, | |
| "loss": 0.8315, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.4311270125223615e-05, | |
| "loss": 0.8359, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.4293381037567084e-05, | |
| "loss": 0.8281, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.4275491949910556e-05, | |
| "loss": 0.8205, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.4257602862254025e-05, | |
| "loss": 0.8177, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.4239713774597497e-05, | |
| "loss": 0.8261, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.4221824686940968e-05, | |
| "loss": 0.8496, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.4203935599284437e-05, | |
| "loss": 0.8056, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.4186046511627909e-05, | |
| "loss": 0.877, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.4168157423971378e-05, | |
| "loss": 0.8444, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.415026833631485e-05, | |
| "loss": 0.8212, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.4132379248658319e-05, | |
| "loss": 0.8268, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.4114490161001791e-05, | |
| "loss": 0.8419, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.409660107334526e-05, | |
| "loss": 0.863, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.407871198568873e-05, | |
| "loss": 0.8222, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.4060822898032202e-05, | |
| "loss": 0.8259, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.4042933810375671e-05, | |
| "loss": 0.7908, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.4025044722719143e-05, | |
| "loss": 0.8292, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.4007155635062612e-05, | |
| "loss": 0.8414, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.3989266547406084e-05, | |
| "loss": 0.8316, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.3971377459749553e-05, | |
| "loss": 0.8278, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.3953488372093025e-05, | |
| "loss": 0.7725, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.3935599284436494e-05, | |
| "loss": 0.8013, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.3917710196779965e-05, | |
| "loss": 0.8475, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.3899821109123435e-05, | |
| "loss": 0.8314, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.3881932021466906e-05, | |
| "loss": 0.8102, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.3864042933810378e-05, | |
| "loss": 0.8208, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.3846153846153847e-05, | |
| "loss": 0.8327, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.3828264758497319e-05, | |
| "loss": 0.8524, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.3810375670840788e-05, | |
| "loss": 0.8201, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.379248658318426e-05, | |
| "loss": 0.8399, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.3774597495527729e-05, | |
| "loss": 0.8894, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.37567084078712e-05, | |
| "loss": 0.8275, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.373881932021467e-05, | |
| "loss": 0.8194, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.372093023255814e-05, | |
| "loss": 0.7857, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.3703041144901612e-05, | |
| "loss": 0.7948, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.3685152057245081e-05, | |
| "loss": 0.832, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.3667262969588553e-05, | |
| "loss": 0.7991, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.3649373881932022e-05, | |
| "loss": 0.8105, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.3631484794275493e-05, | |
| "loss": 0.8577, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.3613595706618963e-05, | |
| "loss": 0.7955, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.3595706618962434e-05, | |
| "loss": 0.827, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.3577817531305904e-05, | |
| "loss": 0.8027, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.3559928443649375e-05, | |
| "loss": 0.8225, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.3542039355992847e-05, | |
| "loss": 0.8167, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.3524150268336316e-05, | |
| "loss": 0.8369, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.3506261180679788e-05, | |
| "loss": 0.7537, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.3488372093023257e-05, | |
| "loss": 0.817, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.3470483005366727e-05, | |
| "loss": 0.7976, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.3452593917710198e-05, | |
| "loss": 0.8305, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.3434704830053668e-05, | |
| "loss": 0.7939, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.3416815742397137e-05, | |
| "loss": 0.8733, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.339892665474061e-05, | |
| "loss": 0.8397, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.3381037567084081e-05, | |
| "loss": 0.7967, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.336314847942755e-05, | |
| "loss": 0.8074, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.334525939177102e-05, | |
| "loss": 0.8251, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.3327370304114491e-05, | |
| "loss": 0.8205, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.3309481216457962e-05, | |
| "loss": 0.8295, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.3291592128801432e-05, | |
| "loss": 0.7885, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.3273703041144903e-05, | |
| "loss": 0.7987, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.3255813953488372e-05, | |
| "loss": 0.7914, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.3237924865831844e-05, | |
| "loss": 0.8191, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.3220035778175316e-05, | |
| "loss": 0.8328, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.3202146690518785e-05, | |
| "loss": 0.7961, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.3184257602862255e-05, | |
| "loss": 0.8636, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.3166368515205726e-05, | |
| "loss": 0.8132, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.3148479427549196e-05, | |
| "loss": 0.7955, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.3130590339892667e-05, | |
| "loss": 0.7777, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.3112701252236137e-05, | |
| "loss": 0.797, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.3094812164579606e-05, | |
| "loss": 0.7756, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.3076923076923078e-05, | |
| "loss": 0.7983, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.3059033989266547e-05, | |
| "loss": 0.8013, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.304114490161002e-05, | |
| "loss": 0.8122, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.302325581395349e-05, | |
| "loss": 0.8039, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.300536672629696e-05, | |
| "loss": 0.791, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.298747763864043e-05, | |
| "loss": 0.7874, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.29695885509839e-05, | |
| "loss": 0.8043, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.2951699463327372e-05, | |
| "loss": 0.848, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.293381037567084e-05, | |
| "loss": 0.8273, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.2915921288014313e-05, | |
| "loss": 0.8292, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.2898032200357782e-05, | |
| "loss": 0.8332, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.2880143112701254e-05, | |
| "loss": 0.8476, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.2862254025044724e-05, | |
| "loss": 0.7687, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.2844364937388195e-05, | |
| "loss": 0.8023, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.2826475849731665e-05, | |
| "loss": 0.8005, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.2808586762075134e-05, | |
| "loss": 0.8098, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.2790697674418606e-05, | |
| "loss": 0.8308, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.2772808586762075e-05, | |
| "loss": 0.8544, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.2754919499105547e-05, | |
| "loss": 0.7893, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.2737030411449016e-05, | |
| "loss": 0.8456, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.2719141323792488e-05, | |
| "loss": 0.8133, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.2701252236135959e-05, | |
| "loss": 0.7946, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.2683363148479428e-05, | |
| "loss": 0.8246, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.26654740608229e-05, | |
| "loss": 0.8294, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.2647584973166369e-05, | |
| "loss": 0.8102, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.262969588550984e-05, | |
| "loss": 0.8149, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.261180679785331e-05, | |
| "loss": 0.8115, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.2593917710196782e-05, | |
| "loss": 0.8004, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.257602862254025e-05, | |
| "loss": 0.8076, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.2558139534883723e-05, | |
| "loss": 0.8292, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.2540250447227193e-05, | |
| "loss": 0.8496, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.2522361359570662e-05, | |
| "loss": 0.7768, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.2504472271914134e-05, | |
| "loss": 0.8446, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.2486583184257603e-05, | |
| "loss": 0.8322, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.2468694096601075e-05, | |
| "loss": 0.7973, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.2450805008944544e-05, | |
| "loss": 0.8085, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.2432915921288016e-05, | |
| "loss": 0.8288, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.2415026833631485e-05, | |
| "loss": 0.8209, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.2397137745974957e-05, | |
| "loss": 0.8287, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.2379248658318426e-05, | |
| "loss": 0.7996, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.2361359570661897e-05, | |
| "loss": 0.7942, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.2343470483005369e-05, | |
| "loss": 0.8388, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.2325581395348838e-05, | |
| "loss": 0.7925, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.230769230769231e-05, | |
| "loss": 0.8203, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.2289803220035779e-05, | |
| "loss": 0.793, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.227191413237925e-05, | |
| "loss": 0.8156, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.225402504472272e-05, | |
| "loss": 0.7924, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.223613595706619e-05, | |
| "loss": 0.8505, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.221824686940966e-05, | |
| "loss": 0.8064, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.2200357781753131e-05, | |
| "loss": 0.7871, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.2182468694096603e-05, | |
| "loss": 0.8574, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.2164579606440072e-05, | |
| "loss": 0.805, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.2146690518783544e-05, | |
| "loss": 0.8176, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.2128801431127013e-05, | |
| "loss": 0.8276, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.2110912343470485e-05, | |
| "loss": 0.7868, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.2093023255813954e-05, | |
| "loss": 0.8524, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.2075134168157425e-05, | |
| "loss": 0.8298, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.2057245080500895e-05, | |
| "loss": 0.8037, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.2039355992844366e-05, | |
| "loss": 0.8241, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.2021466905187838e-05, | |
| "loss": 0.7953, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.2003577817531307e-05, | |
| "loss": 0.8377, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.1985688729874779e-05, | |
| "loss": 0.7755, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.1967799642218248e-05, | |
| "loss": 0.8479, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.194991055456172e-05, | |
| "loss": 0.8316, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.1932021466905189e-05, | |
| "loss": 0.8188, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.1914132379248659e-05, | |
| "loss": 0.7952, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.189624329159213e-05, | |
| "loss": 0.8004, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.18783542039356e-05, | |
| "loss": 0.8277, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.1860465116279072e-05, | |
| "loss": 0.8292, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.1842576028622541e-05, | |
| "loss": 0.8137, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.1824686940966013e-05, | |
| "loss": 0.8006, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.1806797853309482e-05, | |
| "loss": 0.8488, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.1788908765652953e-05, | |
| "loss": 0.802, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.1771019677996423e-05, | |
| "loss": 0.8454, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.1753130590339894e-05, | |
| "loss": 0.8, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.1735241502683364e-05, | |
| "loss": 0.8343, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.1717352415026835e-05, | |
| "loss": 0.8415, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.1699463327370303e-05, | |
| "loss": 0.8016, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.1681574239713776e-05, | |
| "loss": 0.8181, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.1663685152057248e-05, | |
| "loss": 0.8539, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.1645796064400717e-05, | |
| "loss": 0.7782, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.1627906976744187e-05, | |
| "loss": 0.9068, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.1610017889087658e-05, | |
| "loss": 0.8396, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.1592128801431128e-05, | |
| "loss": 0.8077, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.1574239713774597e-05, | |
| "loss": 0.7888, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.1556350626118069e-05, | |
| "loss": 0.8033, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.1538461538461538e-05, | |
| "loss": 0.8127, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.152057245080501e-05, | |
| "loss": 0.804, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.150268336314848e-05, | |
| "loss": 0.799, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.1484794275491951e-05, | |
| "loss": 0.805, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.1466905187835422e-05, | |
| "loss": 0.8184, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.1449016100178892e-05, | |
| "loss": 0.8289, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.1431127012522363e-05, | |
| "loss": 0.795, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.1413237924865831e-05, | |
| "loss": 0.839, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.1395348837209304e-05, | |
| "loss": 0.8114, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.1377459749552772e-05, | |
| "loss": 0.8125, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.1359570661896245e-05, | |
| "loss": 0.7991, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.1341681574239715e-05, | |
| "loss": 0.7785, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.1323792486583186e-05, | |
| "loss": 0.7919, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.1305903398926656e-05, | |
| "loss": 0.8056, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.1288014311270127e-05, | |
| "loss": 0.7898, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.1270125223613597e-05, | |
| "loss": 0.7966, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.1252236135957066e-05, | |
| "loss": 0.8095, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.1234347048300538e-05, | |
| "loss": 0.8608, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.1216457960644007e-05, | |
| "loss": 0.8345, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.1198568872987479e-05, | |
| "loss": 0.8212, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.118067978533095e-05, | |
| "loss": 0.8281, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.116279069767442e-05, | |
| "loss": 0.8135, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.114490161001789e-05, | |
| "loss": 0.8156, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.112701252236136e-05, | |
| "loss": 0.805, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.1109123434704832e-05, | |
| "loss": 0.8174, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.10912343470483e-05, | |
| "loss": 0.7906, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.1073345259391773e-05, | |
| "loss": 0.8287, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.1055456171735241e-05, | |
| "loss": 0.8624, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.1037567084078714e-05, | |
| "loss": 0.8261, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.1019677996422182e-05, | |
| "loss": 0.7726, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.1001788908765655e-05, | |
| "loss": 0.8272, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.0983899821109125e-05, | |
| "loss": 0.8672, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.0966010733452594e-05, | |
| "loss": 0.8423, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.0948121645796066e-05, | |
| "loss": 0.8089, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.0930232558139535e-05, | |
| "loss": 0.8388, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.0912343470483007e-05, | |
| "loss": 0.8015, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.0894454382826476e-05, | |
| "loss": 0.8198, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.0876565295169948e-05, | |
| "loss": 0.7954, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.0858676207513417e-05, | |
| "loss": 0.8267, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.0840787119856887e-05, | |
| "loss": 0.7866, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.082289803220036e-05, | |
| "loss": 0.8219, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.0805008944543828e-05, | |
| "loss": 0.795, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.07871198568873e-05, | |
| "loss": 0.8136, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.076923076923077e-05, | |
| "loss": 0.837, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.0751341681574242e-05, | |
| "loss": 0.8103, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.073345259391771e-05, | |
| "loss": 0.7951, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.0715563506261183e-05, | |
| "loss": 0.831, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.0697674418604651e-05, | |
| "loss": 0.7828, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.0679785330948122e-05, | |
| "loss": 0.816, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.0661896243291594e-05, | |
| "loss": 0.8001, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.0644007155635063e-05, | |
| "loss": 0.8014, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.0626118067978535e-05, | |
| "loss": 0.7981, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.0608228980322004e-05, | |
| "loss": 0.7993, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.0590339892665476e-05, | |
| "loss": 0.8249, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.0572450805008945e-05, | |
| "loss": 0.7812, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.0554561717352417e-05, | |
| "loss": 0.8655, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.0536672629695886e-05, | |
| "loss": 0.8456, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.0518783542039356e-05, | |
| "loss": 0.8372, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.0500894454382829e-05, | |
| "loss": 0.8431, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.0483005366726297e-05, | |
| "loss": 0.844, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.046511627906977e-05, | |
| "loss": 0.8103, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.0447227191413238e-05, | |
| "loss": 0.8066, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.042933810375671e-05, | |
| "loss": 0.8131, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.041144901610018e-05, | |
| "loss": 0.8277, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.039355992844365e-05, | |
| "loss": 0.8115, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.037567084078712e-05, | |
| "loss": 0.8176, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.035778175313059e-05, | |
| "loss": 0.815, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.0339892665474061e-05, | |
| "loss": 0.8207, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.0322003577817532e-05, | |
| "loss": 0.7933, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.0304114490161004e-05, | |
| "loss": 0.8445, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.0286225402504473e-05, | |
| "loss": 0.8, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.0268336314847945e-05, | |
| "loss": 0.8024, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.0250447227191414e-05, | |
| "loss": 0.8016, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.0232558139534884e-05, | |
| "loss": 0.851, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.0214669051878355e-05, | |
| "loss": 0.8425, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.0196779964221825e-05, | |
| "loss": 0.8367, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.0178890876565294e-05, | |
| "loss": 0.7939, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.0161001788908766e-05, | |
| "loss": 0.7758, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.0143112701252239e-05, | |
| "loss": 0.8551, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.0125223613595707e-05, | |
| "loss": 0.806, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.0107334525939178e-05, | |
| "loss": 0.764, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.0089445438282648e-05, | |
| "loss": 0.8303, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.0071556350626119e-05, | |
| "loss": 0.8181, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.005366726296959e-05, | |
| "loss": 0.7926, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.003577817531306e-05, | |
| "loss": 0.812, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.0017889087656529e-05, | |
| "loss": 0.8275, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1e-05, | |
| "loss": 0.7982, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 9.982110912343471e-06, | |
| "loss": 0.803, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 9.964221824686942e-06, | |
| "loss": 0.7827, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 9.946332737030412e-06, | |
| "loss": 0.8194, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 9.928443649373883e-06, | |
| "loss": 0.7769, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.910554561717353e-06, | |
| "loss": 0.8422, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.892665474060824e-06, | |
| "loss": 0.7761, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.874776386404294e-06, | |
| "loss": 0.7761, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.856887298747765e-06, | |
| "loss": 0.7828, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.838998211091235e-06, | |
| "loss": 0.8169, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.821109123434706e-06, | |
| "loss": 0.8082, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.803220035778176e-06, | |
| "loss": 0.8247, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.785330948121647e-06, | |
| "loss": 0.8266, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.767441860465117e-06, | |
| "loss": 0.8293, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 9.749552772808588e-06, | |
| "loss": 0.8246, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 9.731663685152058e-06, | |
| "loss": 0.8179, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 9.713774597495529e-06, | |
| "loss": 0.7831, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 9.695885509839e-06, | |
| "loss": 0.8261, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 9.67799642218247e-06, | |
| "loss": 0.8464, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 9.66010733452594e-06, | |
| "loss": 0.819, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 9.64221824686941e-06, | |
| "loss": 0.8039, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 9.624329159212881e-06, | |
| "loss": 0.8729, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 9.606440071556352e-06, | |
| "loss": 0.8002, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 9.58855098389982e-06, | |
| "loss": 0.8374, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 9.570661896243293e-06, | |
| "loss": 0.7905, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 9.552772808586763e-06, | |
| "loss": 0.8314, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 9.534883720930234e-06, | |
| "loss": 0.8081, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 9.516994633273704e-06, | |
| "loss": 0.7697, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 9.499105545617175e-06, | |
| "loss": 0.8407, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 9.481216457960645e-06, | |
| "loss": 0.7919, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 9.463327370304116e-06, | |
| "loss": 0.8082, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 9.445438282647585e-06, | |
| "loss": 0.8318, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 9.427549194991055e-06, | |
| "loss": 0.775, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 9.409660107334526e-06, | |
| "loss": 0.8003, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 9.391771019677998e-06, | |
| "loss": 0.7851, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 9.373881932021468e-06, | |
| "loss": 0.8138, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 9.355992844364939e-06, | |
| "loss": 0.7826, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 9.33810375670841e-06, | |
| "loss": 0.828, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 9.32021466905188e-06, | |
| "loss": 0.8016, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 9.30232558139535e-06, | |
| "loss": 0.8034, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 9.284436493738819e-06, | |
| "loss": 0.8151, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 9.26654740608229e-06, | |
| "loss": 0.8402, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 9.24865831842576e-06, | |
| "loss": 0.8238, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 9.230769230769232e-06, | |
| "loss": 0.7693, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 9.212880143112703e-06, | |
| "loss": 0.7903, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 9.194991055456173e-06, | |
| "loss": 0.8573, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 9.177101967799644e-06, | |
| "loss": 0.774, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 9.159212880143114e-06, | |
| "loss": 0.8503, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 9.141323792486583e-06, | |
| "loss": 0.8098, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 9.123434704830054e-06, | |
| "loss": 0.8103, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 9.105545617173524e-06, | |
| "loss": 0.7995, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 9.087656529516995e-06, | |
| "loss": 0.8094, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 9.069767441860465e-06, | |
| "loss": 0.8211, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 9.051878354203937e-06, | |
| "loss": 0.8306, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 9.033989266547408e-06, | |
| "loss": 0.7922, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 9.016100178890878e-06, | |
| "loss": 0.8219, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 8.998211091234347e-06, | |
| "loss": 0.8025, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 8.980322003577818e-06, | |
| "loss": 0.8461, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 8.962432915921288e-06, | |
| "loss": 0.7912, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 8.944543828264759e-06, | |
| "loss": 0.7987, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 8.926654740608229e-06, | |
| "loss": 0.8158, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 8.9087656529517e-06, | |
| "loss": 0.78, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 8.890876565295172e-06, | |
| "loss": 0.8035, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 8.872987477638642e-06, | |
| "loss": 0.8186, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 8.855098389982111e-06, | |
| "loss": 0.8076, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 8.837209302325582e-06, | |
| "loss": 0.8235, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 8.819320214669052e-06, | |
| "loss": 0.7922, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 8.801431127012523e-06, | |
| "loss": 0.833, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 8.783542039355993e-06, | |
| "loss": 0.8213, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 8.765652951699464e-06, | |
| "loss": 0.815, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 8.747763864042934e-06, | |
| "loss": 0.8441, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 8.729874776386405e-06, | |
| "loss": 0.7666, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 8.711985688729875e-06, | |
| "loss": 0.8039, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 8.694096601073346e-06, | |
| "loss": 0.8205, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 8.676207513416816e-06, | |
| "loss": 0.8547, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 8.658318425760287e-06, | |
| "loss": 0.8326, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 8.640429338103757e-06, | |
| "loss": 0.8591, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 8.622540250447228e-06, | |
| "loss": 0.7821, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 8.604651162790698e-06, | |
| "loss": 0.785, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 8.586762075134169e-06, | |
| "loss": 0.7872, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 8.568872987477639e-06, | |
| "loss": 0.8151, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 8.55098389982111e-06, | |
| "loss": 0.8094, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 8.53309481216458e-06, | |
| "loss": 0.8497, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 8.51520572450805e-06, | |
| "loss": 0.817, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 8.497316636851521e-06, | |
| "loss": 0.7746, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 8.479427549194992e-06, | |
| "loss": 0.7654, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 8.461538461538462e-06, | |
| "loss": 0.8039, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 8.443649373881933e-06, | |
| "loss": 0.7995, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 8.425760286225403e-06, | |
| "loss": 0.7986, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 8.407871198568874e-06, | |
| "loss": 0.7987, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 8.389982110912344e-06, | |
| "loss": 0.8235, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 8.372093023255815e-06, | |
| "loss": 0.8139, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 8.354203935599285e-06, | |
| "loss": 0.7667, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 8.336314847942756e-06, | |
| "loss": 0.763, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 8.318425760286226e-06, | |
| "loss": 0.8184, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 8.300536672629697e-06, | |
| "loss": 0.8314, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 8.282647584973167e-06, | |
| "loss": 0.7943, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 8.264758497316638e-06, | |
| "loss": 0.8006, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 8.246869409660108e-06, | |
| "loss": 0.7979, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 8.228980322003579e-06, | |
| "loss": 0.7878, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 8.211091234347049e-06, | |
| "loss": 0.7726, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 8.19320214669052e-06, | |
| "loss": 0.7806, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 8.17531305903399e-06, | |
| "loss": 0.8422, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 8.15742397137746e-06, | |
| "loss": 0.8249, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 8.139534883720931e-06, | |
| "loss": 0.7869, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 8.121645796064402e-06, | |
| "loss": 0.7759, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 8.103756708407872e-06, | |
| "loss": 0.832, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 8.085867620751343e-06, | |
| "loss": 0.8425, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 8.067978533094813e-06, | |
| "loss": 0.7577, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 8.050089445438284e-06, | |
| "loss": 0.7994, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 8.032200357781754e-06, | |
| "loss": 0.7971, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 8.014311270125225e-06, | |
| "loss": 0.8123, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 7.996422182468695e-06, | |
| "loss": 0.8269, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 7.978533094812166e-06, | |
| "loss": 0.8499, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 7.960644007155636e-06, | |
| "loss": 0.8749, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 7.942754919499107e-06, | |
| "loss": 0.7885, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 7.924865831842577e-06, | |
| "loss": 0.8209, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 7.906976744186048e-06, | |
| "loss": 0.7713, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 7.889087656529516e-06, | |
| "loss": 0.7758, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 7.871198568872989e-06, | |
| "loss": 0.7752, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 7.853309481216459e-06, | |
| "loss": 0.7854, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 7.83542039355993e-06, | |
| "loss": 0.8368, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 7.8175313059034e-06, | |
| "loss": 0.8165, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 7.79964221824687e-06, | |
| "loss": 0.8105, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 7.781753130590341e-06, | |
| "loss": 0.8101, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 7.763864042933812e-06, | |
| "loss": 0.7787, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 7.74597495527728e-06, | |
| "loss": 0.8178, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 7.728085867620751e-06, | |
| "loss": 0.8253, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 7.710196779964223e-06, | |
| "loss": 0.7993, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 7.692307692307694e-06, | |
| "loss": 0.8172, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 7.674418604651164e-06, | |
| "loss": 0.837, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 7.656529516994635e-06, | |
| "loss": 0.8311, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 7.638640429338105e-06, | |
| "loss": 0.837, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 7.620751341681575e-06, | |
| "loss": 0.8171, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 7.602862254025045e-06, | |
| "loss": 0.8511, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 7.584973166368516e-06, | |
| "loss": 0.8214, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 7.567084078711986e-06, | |
| "loss": 0.8072, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 7.549194991055456e-06, | |
| "loss": 0.7995, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 7.531305903398928e-06, | |
| "loss": 0.8408, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 7.513416815742398e-06, | |
| "loss": 0.7955, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 7.495527728085868e-06, | |
| "loss": 0.8309, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 7.477638640429339e-06, | |
| "loss": 0.7957, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 7.459749552772809e-06, | |
| "loss": 0.8013, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 7.44186046511628e-06, | |
| "loss": 0.7974, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 7.42397137745975e-06, | |
| "loss": 0.8221, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 7.40608228980322e-06, | |
| "loss": 0.8173, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 7.38819320214669e-06, | |
| "loss": 0.8378, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 7.370304114490162e-06, | |
| "loss": 0.8092, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 7.352415026833632e-06, | |
| "loss": 0.7766, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 7.334525939177103e-06, | |
| "loss": 0.8171, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 7.316636851520573e-06, | |
| "loss": 0.7753, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 7.298747763864044e-06, | |
| "loss": 0.8408, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 7.280858676207514e-06, | |
| "loss": 0.8039, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 7.262969588550985e-06, | |
| "loss": 0.8069, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 7.245080500894454e-06, | |
| "loss": 0.9911, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 7.227191413237925e-06, | |
| "loss": 0.811, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 7.209302325581395e-06, | |
| "loss": 0.7949, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 7.191413237924867e-06, | |
| "loss": 0.7854, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 7.173524150268337e-06, | |
| "loss": 0.8311, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 7.155635062611808e-06, | |
| "loss": 0.8619, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 7.137745974955278e-06, | |
| "loss": 0.8464, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 7.119856887298749e-06, | |
| "loss": 0.7921, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 7.101967799642218e-06, | |
| "loss": 0.805, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 7.084078711985689e-06, | |
| "loss": 0.7765, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 7.066189624329159e-06, | |
| "loss": 0.8042, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 7.04830053667263e-06, | |
| "loss": 0.8153, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 7.030411449016101e-06, | |
| "loss": 0.7911, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 7.012522361359572e-06, | |
| "loss": 0.7922, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 6.994633273703042e-06, | |
| "loss": 0.7909, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 6.976744186046513e-06, | |
| "loss": 0.7952, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 6.958855098389982e-06, | |
| "loss": 0.8167, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 6.940966010733453e-06, | |
| "loss": 0.8124, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 6.923076923076923e-06, | |
| "loss": 0.8123, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 6.905187835420394e-06, | |
| "loss": 0.8333, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 6.887298747763864e-06, | |
| "loss": 0.7883, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 6.869409660107335e-06, | |
| "loss": 0.8433, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 6.851520572450806e-06, | |
| "loss": 0.7817, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 6.833631484794277e-06, | |
| "loss": 0.8395, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 6.815742397137746e-06, | |
| "loss": 0.7844, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 6.797853309481217e-06, | |
| "loss": 0.8647, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 6.779964221824687e-06, | |
| "loss": 0.821, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 6.762075134168158e-06, | |
| "loss": 0.8183, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 6.744186046511628e-06, | |
| "loss": 0.8236, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 6.726296958855099e-06, | |
| "loss": 0.806, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 6.7084078711985685e-06, | |
| "loss": 0.8363, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 6.690518783542041e-06, | |
| "loss": 0.7865, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 6.67262969588551e-06, | |
| "loss": 0.7991, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 6.654740608228981e-06, | |
| "loss": 0.8444, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 6.636851520572451e-06, | |
| "loss": 0.83, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 6.618962432915922e-06, | |
| "loss": 0.7788, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 6.601073345259392e-06, | |
| "loss": 0.7715, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 6.583184257602863e-06, | |
| "loss": 0.8047, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 6.565295169946333e-06, | |
| "loss": 0.7993, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 6.547406082289803e-06, | |
| "loss": 0.7604, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 6.5295169946332735e-06, | |
| "loss": 0.8024, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 6.511627906976745e-06, | |
| "loss": 0.8321, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 6.493738819320215e-06, | |
| "loss": 0.7693, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 6.475849731663686e-06, | |
| "loss": 0.7734, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 6.457960644007156e-06, | |
| "loss": 0.7792, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 6.440071556350627e-06, | |
| "loss": 0.8107, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 6.422182468694097e-06, | |
| "loss": 0.8167, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 6.404293381037567e-06, | |
| "loss": 0.8422, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 6.3864042933810375e-06, | |
| "loss": 0.8242, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 6.368515205724508e-06, | |
| "loss": 0.7872, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 6.350626118067979e-06, | |
| "loss": 0.8181, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 6.33273703041145e-06, | |
| "loss": 0.8122, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 6.31484794275492e-06, | |
| "loss": 0.8193, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 6.296958855098391e-06, | |
| "loss": 0.8085, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 6.279069767441861e-06, | |
| "loss": 0.7717, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 6.261180679785331e-06, | |
| "loss": 0.8203, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 6.2432915921288015e-06, | |
| "loss": 0.8146, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 6.225402504472272e-06, | |
| "loss": 0.8133, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 6.2075134168157425e-06, | |
| "loss": 0.8394, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 6.189624329159213e-06, | |
| "loss": 0.7894, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 6.171735241502684e-06, | |
| "loss": 0.7813, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 6.153846153846155e-06, | |
| "loss": 0.805, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 6.135957066189625e-06, | |
| "loss": 0.7934, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 6.118067978533095e-06, | |
| "loss": 0.773, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 6.1001788908765655e-06, | |
| "loss": 0.778, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 6.082289803220036e-06, | |
| "loss": 0.8251, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 6.0644007155635065e-06, | |
| "loss": 0.7945, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 6.046511627906977e-06, | |
| "loss": 0.7968, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 6.0286225402504475e-06, | |
| "loss": 0.8071, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 6.010733452593919e-06, | |
| "loss": 0.8697, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 5.992844364937389e-06, | |
| "loss": 0.7932, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 5.97495527728086e-06, | |
| "loss": 0.7877, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 5.9570661896243295e-06, | |
| "loss": 0.8736, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 5.9391771019678e-06, | |
| "loss": 0.8103, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 5.9212880143112705e-06, | |
| "loss": 0.7899, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 5.903398926654741e-06, | |
| "loss": 0.7992, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 5.8855098389982115e-06, | |
| "loss": 0.7849, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 5.867620751341682e-06, | |
| "loss": 0.8108, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 5.849731663685152e-06, | |
| "loss": 0.7346, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 5.831842576028624e-06, | |
| "loss": 0.8016, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 5.8139534883720935e-06, | |
| "loss": 0.8069, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 5.796064400715564e-06, | |
| "loss": 0.7909, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 5.7781753130590345e-06, | |
| "loss": 0.8149, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 5.760286225402505e-06, | |
| "loss": 0.9054, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 5.7423971377459755e-06, | |
| "loss": 0.8417, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 5.724508050089446e-06, | |
| "loss": 0.8227, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 5.706618962432916e-06, | |
| "loss": 0.7947, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 5.688729874776386e-06, | |
| "loss": 0.8325, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 5.6708407871198575e-06, | |
| "loss": 0.8571, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 5.652951699463328e-06, | |
| "loss": 0.7707, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 5.6350626118067985e-06, | |
| "loss": 0.8276, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 5.617173524150269e-06, | |
| "loss": 0.8805, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 5.5992844364937395e-06, | |
| "loss": 0.7982, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 5.58139534883721e-06, | |
| "loss": 0.7948, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 5.56350626118068e-06, | |
| "loss": 0.7969, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 5.54561717352415e-06, | |
| "loss": 0.7887, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 5.527728085867621e-06, | |
| "loss": 0.8008, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 5.509838998211091e-06, | |
| "loss": 0.8276, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 5.4919499105545625e-06, | |
| "loss": 0.754, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 5.474060822898033e-06, | |
| "loss": 0.8439, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 5.4561717352415035e-06, | |
| "loss": 0.8363, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 5.438282647584974e-06, | |
| "loss": 0.7987, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 5.420393559928444e-06, | |
| "loss": 0.7765, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 5.402504472271914e-06, | |
| "loss": 0.7706, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 5.384615384615385e-06, | |
| "loss": 0.8249, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 5.366726296958855e-06, | |
| "loss": 0.771, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 5.348837209302326e-06, | |
| "loss": 0.8427, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 5.330948121645797e-06, | |
| "loss": 0.8006, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 5.3130590339892675e-06, | |
| "loss": 0.7935, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 5.295169946332738e-06, | |
| "loss": 0.8035, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 5.2772808586762085e-06, | |
| "loss": 0.7927, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 5.259391771019678e-06, | |
| "loss": 0.8362, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 5.241502683363149e-06, | |
| "loss": 0.8026, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 5.223613595706619e-06, | |
| "loss": 0.8325, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 5.20572450805009e-06, | |
| "loss": 0.7672, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 5.18783542039356e-06, | |
| "loss": 0.8045, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 5.169946332737031e-06, | |
| "loss": 0.7638, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 5.152057245080502e-06, | |
| "loss": 0.7772, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 5.1341681574239725e-06, | |
| "loss": 0.8317, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 5.116279069767442e-06, | |
| "loss": 0.7794, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 5.098389982110913e-06, | |
| "loss": 0.8264, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 5.080500894454383e-06, | |
| "loss": 0.7952, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 5.062611806797854e-06, | |
| "loss": 0.7527, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 5.044722719141324e-06, | |
| "loss": 0.8192, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 5.026833631484795e-06, | |
| "loss": 0.8254, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 5.008944543828264e-06, | |
| "loss": 0.8244, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 4.991055456171736e-06, | |
| "loss": 0.8175, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 4.973166368515206e-06, | |
| "loss": 0.8086, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 4.955277280858677e-06, | |
| "loss": 0.8022, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 4.937388193202147e-06, | |
| "loss": 0.8121, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 4.919499105545618e-06, | |
| "loss": 0.8112, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 4.901610017889088e-06, | |
| "loss": 0.795, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 4.883720930232559e-06, | |
| "loss": 0.7895, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 4.865831842576029e-06, | |
| "loss": 0.801, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 4.8479427549195e-06, | |
| "loss": 0.8463, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 4.83005366726297e-06, | |
| "loss": 0.7875, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.812164579606441e-06, | |
| "loss": 0.8168, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.79427549194991e-06, | |
| "loss": 0.7904, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.776386404293382e-06, | |
| "loss": 0.8368, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.758497316636852e-06, | |
| "loss": 0.7661, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 4.740608228980323e-06, | |
| "loss": 0.8313, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 4.722719141323792e-06, | |
| "loss": 0.8063, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 4.704830053667263e-06, | |
| "loss": 0.7873, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 4.686940966010734e-06, | |
| "loss": 0.8313, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 4.669051878354205e-06, | |
| "loss": 0.7929, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 4.651162790697675e-06, | |
| "loss": 0.7602, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 4.633273703041145e-06, | |
| "loss": 0.7963, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 4.615384615384616e-06, | |
| "loss": 0.7846, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 4.597495527728087e-06, | |
| "loss": 0.7958, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 4.579606440071557e-06, | |
| "loss": 0.8215, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 4.561717352415027e-06, | |
| "loss": 0.7864, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 4.543828264758497e-06, | |
| "loss": 0.798, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 4.525939177101969e-06, | |
| "loss": 0.8171, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 4.508050089445439e-06, | |
| "loss": 0.7894, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 4.490161001788909e-06, | |
| "loss": 0.7753, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 4.472271914132379e-06, | |
| "loss": 0.7605, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 4.45438282647585e-06, | |
| "loss": 0.7998, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 4.436493738819321e-06, | |
| "loss": 0.789, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 4.418604651162791e-06, | |
| "loss": 0.8018, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 4.400715563506261e-06, | |
| "loss": 0.7884, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 4.382826475849732e-06, | |
| "loss": 0.7946, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 4.364937388193202e-06, | |
| "loss": 0.821, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 4.347048300536673e-06, | |
| "loss": 0.8094, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 4.329159212880143e-06, | |
| "loss": 0.7607, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 4.311270125223614e-06, | |
| "loss": 0.8227, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 4.293381037567084e-06, | |
| "loss": 0.8209, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 4.275491949910555e-06, | |
| "loss": 0.8149, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 4.257602862254025e-06, | |
| "loss": 0.8101, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 4.239713774597496e-06, | |
| "loss": 0.8114, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 4.221824686940966e-06, | |
| "loss": 0.8464, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 4.203935599284437e-06, | |
| "loss": 0.7864, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 4.186046511627907e-06, | |
| "loss": 0.7908, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.168157423971378e-06, | |
| "loss": 0.7498, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.150268336314848e-06, | |
| "loss": 0.7956, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.132379248658319e-06, | |
| "loss": 0.7902, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.114490161001789e-06, | |
| "loss": 0.8174, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 4.09660107334526e-06, | |
| "loss": 0.7848, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 4.07871198568873e-06, | |
| "loss": 0.7947, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 4.060822898032201e-06, | |
| "loss": 0.7701, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 4.042933810375671e-06, | |
| "loss": 0.8441, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 4.025044722719142e-06, | |
| "loss": 0.7954, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 4.007155635062612e-06, | |
| "loss": 0.8064, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 3.989266547406083e-06, | |
| "loss": 0.8712, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 3.971377459749553e-06, | |
| "loss": 0.7821, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 3.953488372093024e-06, | |
| "loss": 0.8187, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 3.935599284436494e-06, | |
| "loss": 0.828, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 3.917710196779965e-06, | |
| "loss": 0.8434, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 3.899821109123435e-06, | |
| "loss": 0.7795, | |
| "step": 1000 | |
| } | |
| ], | |
| "max_steps": 1218, | |
| "num_train_epochs": 3, | |
| "total_flos": 1.30015927338428e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |