| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 2.55863539445629, |
| "eval_steps": 500, |
| "global_step": 1200, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.0, |
| "loss": 15.3957, |
| "step": 1 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 1.0905214599276578e-06, |
| "loss": 15.2565, |
| "step": 2 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 1.7284356202170269e-06, |
| "loss": 15.3225, |
| "step": 3 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.1810429198553156e-06, |
| "loss": 15.0692, |
| "step": 4 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.5321124158836113e-06, |
| "loss": 14.9505, |
| "step": 5 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.818957080144684e-06, |
| "loss": 14.215, |
| "step": 6 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 3.0614807881373545e-06, |
| "loss": 14.1495, |
| "step": 7 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 3.2715643797829733e-06, |
| "loss": 13.832, |
| "step": 8 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 3.4568712404340537e-06, |
| "loss": 11.9035, |
| "step": 9 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 3.62263387581127e-06, |
| "loss": 11.6671, |
| "step": 10 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 3.7725844192762463e-06, |
| "loss": 11.3183, |
| "step": 11 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 3.909478540072343e-06, |
| "loss": 10.8698, |
| "step": 12 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 4.035408923801514e-06, |
| "loss": 10.6018, |
| "step": 13 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 4.152002248065012e-06, |
| "loss": 6.4661, |
| "step": 14 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 4.260548036100639e-06, |
| "loss": 5.862, |
| "step": 15 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 4.362085839710631e-06, |
| "loss": 5.122, |
| "step": 16 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.457465945040373e-06, |
| "loss": 4.0585, |
| "step": 17 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.547392700361711e-06, |
| "loss": 3.0055, |
| "step": 18 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.632456113627364e-06, |
| "loss": 2.1107, |
| "step": 19 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.713155335738927e-06, |
| "loss": 1.5126, |
| "step": 20 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 4.7899164083543815e-06, |
| "loss": 0.7493, |
| "step": 21 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 4.863105879203904e-06, |
| "loss": 0.2574, |
| "step": 22 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 4.9330413883925056e-06, |
| "loss": 0.111, |
| "step": 23 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 5e-06, |
| "loss": 0.2328, |
| "step": 24 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 5e-06, |
| "loss": 0.1597, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 5e-06, |
| "loss": 0.1277, |
| "step": 26 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 5e-06, |
| "loss": 0.078, |
| "step": 27 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 5e-06, |
| "loss": 0.2067, |
| "step": 28 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 5e-06, |
| "loss": 0.2154, |
| "step": 29 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 5e-06, |
| "loss": 0.061, |
| "step": 30 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 5e-06, |
| "loss": 0.1374, |
| "step": 31 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 5e-06, |
| "loss": 0.1866, |
| "step": 32 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 5e-06, |
| "loss": 0.1758, |
| "step": 33 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0699, |
| "step": 34 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 5e-06, |
| "loss": 0.2269, |
| "step": 35 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 5e-06, |
| "loss": 0.3749, |
| "step": 36 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 5e-06, |
| "loss": 0.3111, |
| "step": 37 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 5e-06, |
| "loss": 0.1199, |
| "step": 38 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 5e-06, |
| "loss": 0.093, |
| "step": 39 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0629, |
| "step": 40 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0785, |
| "step": 41 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0573, |
| "step": 42 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0629, |
| "step": 43 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0521, |
| "step": 44 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0578, |
| "step": 45 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0443, |
| "step": 46 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 5e-06, |
| "loss": 0.047, |
| "step": 47 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0614, |
| "step": 48 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0574, |
| "step": 49 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 5e-06, |
| "loss": 0.051, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0418, |
| "step": 51 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0609, |
| "step": 52 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0585, |
| "step": 53 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0945, |
| "step": 54 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 5e-06, |
| "loss": 0.1248, |
| "step": 55 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0641, |
| "step": 56 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0842, |
| "step": 57 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 5e-06, |
| "loss": 0.1111, |
| "step": 58 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0957, |
| "step": 59 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0635, |
| "step": 60 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0624, |
| "step": 61 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0543, |
| "step": 62 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0607, |
| "step": 63 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0461, |
| "step": 64 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0685, |
| "step": 65 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0592, |
| "step": 66 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0512, |
| "step": 67 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 5e-06, |
| "loss": 0.057, |
| "step": 68 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0736, |
| "step": 69 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0542, |
| "step": 70 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0458, |
| "step": 71 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0516, |
| "step": 72 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0451, |
| "step": 73 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0436, |
| "step": 74 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0443, |
| "step": 75 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0559, |
| "step": 76 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0534, |
| "step": 77 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0396, |
| "step": 78 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0522, |
| "step": 79 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 5e-06, |
| "loss": 0.047, |
| "step": 80 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0502, |
| "step": 81 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0475, |
| "step": 82 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0753, |
| "step": 83 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0736, |
| "step": 84 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0519, |
| "step": 85 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0538, |
| "step": 86 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0561, |
| "step": 87 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0532, |
| "step": 88 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0563, |
| "step": 89 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0616, |
| "step": 90 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0477, |
| "step": 91 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0495, |
| "step": 92 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0534, |
| "step": 93 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0562, |
| "step": 94 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0376, |
| "step": 95 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0676, |
| "step": 96 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0505, |
| "step": 97 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0509, |
| "step": 98 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0713, |
| "step": 99 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0553, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0701, |
| "step": 101 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0422, |
| "step": 102 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0401, |
| "step": 103 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0619, |
| "step": 104 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0284, |
| "step": 105 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0405, |
| "step": 106 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0555, |
| "step": 107 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0349, |
| "step": 108 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0495, |
| "step": 109 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0268, |
| "step": 110 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0398, |
| "step": 111 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0352, |
| "step": 112 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0245, |
| "step": 113 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0208, |
| "step": 114 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0318, |
| "step": 115 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0268, |
| "step": 116 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0237, |
| "step": 117 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0274, |
| "step": 118 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0173, |
| "step": 119 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0429, |
| "step": 120 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0292, |
| "step": 121 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0362, |
| "step": 122 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0165, |
| "step": 123 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0189, |
| "step": 124 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 5e-06, |
| "loss": 0.015, |
| "step": 125 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0186, |
| "step": 126 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0235, |
| "step": 127 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0247, |
| "step": 128 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0233, |
| "step": 129 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0192, |
| "step": 130 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0174, |
| "step": 131 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 5e-06, |
| "loss": 0.03, |
| "step": 132 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 133 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0199, |
| "step": 134 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 135 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0253, |
| "step": 136 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 137 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0262, |
| "step": 138 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0154, |
| "step": 139 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0201, |
| "step": 140 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0109, |
| "step": 141 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0142, |
| "step": 142 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0191, |
| "step": 143 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0283, |
| "step": 144 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0305, |
| "step": 145 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0215, |
| "step": 146 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0131, |
| "step": 147 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0225, |
| "step": 148 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0149, |
| "step": 149 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0167, |
| "step": 150 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0302, |
| "step": 151 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0198, |
| "step": 152 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0188, |
| "step": 153 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0161, |
| "step": 154 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0256, |
| "step": 155 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0301, |
| "step": 156 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0137, |
| "step": 157 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0337, |
| "step": 158 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0209, |
| "step": 159 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0121, |
| "step": 160 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0183, |
| "step": 161 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0163, |
| "step": 162 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0174, |
| "step": 163 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0167, |
| "step": 164 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0191, |
| "step": 165 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0188, |
| "step": 166 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0115, |
| "step": 167 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0228, |
| "step": 168 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0136, |
| "step": 169 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0266, |
| "step": 170 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0296, |
| "step": 171 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0206, |
| "step": 172 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 5e-06, |
| "loss": 0.014, |
| "step": 173 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 174 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0082, |
| "step": 175 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0255, |
| "step": 176 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0088, |
| "step": 177 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0132, |
| "step": 178 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0259, |
| "step": 179 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0098, |
| "step": 180 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0135, |
| "step": 181 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0297, |
| "step": 182 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0203, |
| "step": 183 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0256, |
| "step": 184 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0197, |
| "step": 185 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0065, |
| "step": 186 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0214, |
| "step": 187 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0151, |
| "step": 188 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0217, |
| "step": 189 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 5e-06, |
| "loss": 0.029, |
| "step": 190 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0198, |
| "step": 191 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0186, |
| "step": 192 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0244, |
| "step": 193 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 5e-06, |
| "loss": 0.017, |
| "step": 194 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5e-06, |
| "loss": 0.022, |
| "step": 195 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0073, |
| "step": 196 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0173, |
| "step": 197 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 198 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0088, |
| "step": 199 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0171, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0264, |
| "step": 201 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0259, |
| "step": 202 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0155, |
| "step": 203 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0191, |
| "step": 204 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5e-06, |
| "loss": 0.016, |
| "step": 205 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0242, |
| "step": 206 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0332, |
| "step": 207 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0327, |
| "step": 208 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0231, |
| "step": 209 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0122, |
| "step": 210 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0245, |
| "step": 211 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0194, |
| "step": 212 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 5e-06, |
| "loss": 0.017, |
| "step": 213 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5e-06, |
| "loss": 0.02, |
| "step": 214 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0373, |
| "step": 215 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0233, |
| "step": 216 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0439, |
| "step": 217 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0087, |
| "step": 218 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0095, |
| "step": 219 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0176, |
| "step": 220 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0142, |
| "step": 221 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0088, |
| "step": 222 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0216, |
| "step": 223 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0221, |
| "step": 224 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0173, |
| "step": 225 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0114, |
| "step": 226 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5e-06, |
| "loss": 0.021, |
| "step": 227 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5e-06, |
| "loss": 0.013, |
| "step": 228 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0231, |
| "step": 229 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5e-06, |
| "loss": 0.007, |
| "step": 230 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0114, |
| "step": 231 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 232 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0301, |
| "step": 233 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0296, |
| "step": 234 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0084, |
| "step": 235 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0158, |
| "step": 236 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0158, |
| "step": 237 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5e-06, |
| "loss": 0.024, |
| "step": 238 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0052, |
| "step": 239 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0101, |
| "step": 240 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5e-06, |
| "loss": 0.044, |
| "step": 241 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0259, |
| "step": 242 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0424, |
| "step": 243 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 244 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 245 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0132, |
| "step": 246 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0325, |
| "step": 247 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 248 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0307, |
| "step": 249 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0218, |
| "step": 250 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0293, |
| "step": 251 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0273, |
| "step": 252 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0098, |
| "step": 253 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 5e-06, |
| "loss": 0.02, |
| "step": 254 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0135, |
| "step": 255 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0185, |
| "step": 256 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 257 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0144, |
| "step": 258 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 259 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0094, |
| "step": 260 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0187, |
| "step": 261 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 5e-06, |
| "loss": 0.016, |
| "step": 262 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0213, |
| "step": 263 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0178, |
| "step": 264 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 265 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0125, |
| "step": 266 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0081, |
| "step": 267 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0077, |
| "step": 268 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0129, |
| "step": 269 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0037, |
| "step": 270 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0124, |
| "step": 271 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0231, |
| "step": 272 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0221, |
| "step": 273 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0336, |
| "step": 274 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 5e-06, |
| "loss": 0.008, |
| "step": 275 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 276 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0146, |
| "step": 277 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0125, |
| "step": 278 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0123, |
| "step": 279 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 5e-06, |
| "loss": 0.0118, |
| "step": 280 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 5e-06, |
| "loss": 0.013, |
| "step": 281 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 5e-06, |
| "loss": 0.0169, |
| "step": 282 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 5e-06, |
| "loss": 0.009, |
| "step": 283 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 5e-06, |
| "loss": 0.019, |
| "step": 284 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 5e-06, |
| "loss": 0.0139, |
| "step": 285 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 286 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 5e-06, |
| "loss": 0.0088, |
| "step": 287 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 5e-06, |
| "loss": 0.017, |
| "step": 288 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 5e-06, |
| "loss": 0.0033, |
| "step": 289 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 5e-06, |
| "loss": 0.0127, |
| "step": 290 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 5e-06, |
| "loss": 0.0165, |
| "step": 291 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 5e-06, |
| "loss": 0.0209, |
| "step": 292 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 293 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 5e-06, |
| "loss": 0.008, |
| "step": 294 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 5e-06, |
| "loss": 0.0179, |
| "step": 295 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 5e-06, |
| "loss": 0.029, |
| "step": 296 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 5e-06, |
| "loss": 0.0168, |
| "step": 297 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 5e-06, |
| "loss": 0.012, |
| "step": 298 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 5e-06, |
| "loss": 0.0158, |
| "step": 299 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 5e-06, |
| "loss": 0.0082, |
| "step": 300 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 5e-06, |
| "loss": 0.0121, |
| "step": 301 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 5e-06, |
| "loss": 0.0044, |
| "step": 302 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0148, |
| "step": 303 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0239, |
| "step": 304 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 5e-06, |
| "loss": 0.026, |
| "step": 305 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0075, |
| "step": 306 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0068, |
| "step": 307 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 5e-06, |
| "loss": 0.0103, |
| "step": 308 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 5e-06, |
| "loss": 0.0112, |
| "step": 309 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 5e-06, |
| "loss": 0.0085, |
| "step": 310 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 5e-06, |
| "loss": 0.0172, |
| "step": 311 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 5e-06, |
| "loss": 0.0126, |
| "step": 312 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 313 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 314 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 5e-06, |
| "loss": 0.0181, |
| "step": 315 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 5e-06, |
| "loss": 0.0213, |
| "step": 316 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0099, |
| "step": 317 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 318 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0169, |
| "step": 319 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0219, |
| "step": 320 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0147, |
| "step": 321 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 5e-06, |
| "loss": 0.0046, |
| "step": 322 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 5e-06, |
| "loss": 0.0098, |
| "step": 323 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 5e-06, |
| "loss": 0.0092, |
| "step": 324 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 5e-06, |
| "loss": 0.0113, |
| "step": 325 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0158, |
| "step": 326 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 5e-06, |
| "loss": 0.011, |
| "step": 327 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0118, |
| "step": 328 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0142, |
| "step": 329 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 330 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0152, |
| "step": 331 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0147, |
| "step": 332 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 333 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0112, |
| "step": 334 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 335 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 336 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0059, |
| "step": 337 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0107, |
| "step": 338 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0143, |
| "step": 339 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 5e-06, |
| "loss": 0.016, |
| "step": 340 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 5e-06, |
| "loss": 0.0158, |
| "step": 341 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 5e-06, |
| "loss": 0.0198, |
| "step": 342 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 5e-06, |
| "loss": 0.0108, |
| "step": 343 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 5e-06, |
| "loss": 0.0242, |
| "step": 344 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0183, |
| "step": 345 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0197, |
| "step": 346 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0127, |
| "step": 347 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0175, |
| "step": 348 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0147, |
| "step": 349 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0155, |
| "step": 350 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0178, |
| "step": 351 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 5e-06, |
| "loss": 0.013, |
| "step": 352 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0146, |
| "step": 353 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0223, |
| "step": 354 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 5e-06, |
| "loss": 0.0201, |
| "step": 355 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 5e-06, |
| "loss": 0.0077, |
| "step": 356 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 5e-06, |
| "loss": 0.0142, |
| "step": 357 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 5e-06, |
| "loss": 0.0136, |
| "step": 358 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0068, |
| "step": 359 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0163, |
| "step": 360 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0041, |
| "step": 361 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 5e-06, |
| "loss": 0.016, |
| "step": 362 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0264, |
| "step": 363 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0235, |
| "step": 364 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0186, |
| "step": 365 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0028, |
| "step": 366 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0052, |
| "step": 367 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0086, |
| "step": 368 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 5e-06, |
| "loss": 0.0303, |
| "step": 369 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 5e-06, |
| "loss": 0.0333, |
| "step": 370 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 5e-06, |
| "loss": 0.0291, |
| "step": 371 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 5e-06, |
| "loss": 0.0045, |
| "step": 372 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 373 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0294, |
| "step": 374 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 5e-06, |
| "loss": 0.019, |
| "step": 375 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0166, |
| "step": 376 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 377 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 5e-06, |
| "loss": 0.0141, |
| "step": 378 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 5e-06, |
| "loss": 0.0123, |
| "step": 379 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 5e-06, |
| "loss": 0.0153, |
| "step": 380 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 5e-06, |
| "loss": 0.0235, |
| "step": 381 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 5e-06, |
| "loss": 0.0123, |
| "step": 382 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 5e-06, |
| "loss": 0.0215, |
| "step": 383 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 5e-06, |
| "loss": 0.0115, |
| "step": 384 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 385 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 5e-06, |
| "loss": 0.015, |
| "step": 386 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0174, |
| "step": 387 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0273, |
| "step": 388 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0147, |
| "step": 389 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0114, |
| "step": 390 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0048, |
| "step": 391 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0066, |
| "step": 392 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 393 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0132, |
| "step": 394 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0328, |
| "step": 395 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0041, |
| "step": 396 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 5e-06, |
| "loss": 0.0128, |
| "step": 397 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 5e-06, |
| "loss": 0.0121, |
| "step": 398 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 5e-06, |
| "loss": 0.0216, |
| "step": 399 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 5e-06, |
| "loss": 0.0268, |
| "step": 400 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 5e-06, |
| "loss": 0.015, |
| "step": 401 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 5e-06, |
| "loss": 0.0092, |
| "step": 402 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 5e-06, |
| "loss": 0.009, |
| "step": 403 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 5e-06, |
| "loss": 0.0128, |
| "step": 404 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 5e-06, |
| "loss": 0.005, |
| "step": 405 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 5e-06, |
| "loss": 0.0079, |
| "step": 406 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 5e-06, |
| "loss": 0.009, |
| "step": 407 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 5e-06, |
| "loss": 0.0326, |
| "step": 408 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 5e-06, |
| "loss": 0.0162, |
| "step": 409 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 5e-06, |
| "loss": 0.0055, |
| "step": 410 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 5e-06, |
| "loss": 0.014, |
| "step": 411 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 5e-06, |
| "loss": 0.0046, |
| "step": 412 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 5e-06, |
| "loss": 0.0121, |
| "step": 413 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 5e-06, |
| "loss": 0.0095, |
| "step": 414 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 5e-06, |
| "loss": 0.0106, |
| "step": 415 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 5e-06, |
| "loss": 0.0176, |
| "step": 416 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 5e-06, |
| "loss": 0.0288, |
| "step": 417 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 5e-06, |
| "loss": 0.0128, |
| "step": 418 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 5e-06, |
| "loss": 0.0052, |
| "step": 419 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 5e-06, |
| "loss": 0.0262, |
| "step": 420 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 5e-06, |
| "loss": 0.0106, |
| "step": 421 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 5e-06, |
| "loss": 0.0219, |
| "step": 422 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 5e-06, |
| "loss": 0.0258, |
| "step": 423 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 5e-06, |
| "loss": 0.0079, |
| "step": 424 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 5e-06, |
| "loss": 0.013, |
| "step": 425 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 5e-06, |
| "loss": 0.0216, |
| "step": 426 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 5e-06, |
| "loss": 0.0189, |
| "step": 427 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 5e-06, |
| "loss": 0.0209, |
| "step": 428 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 5e-06, |
| "loss": 0.0234, |
| "step": 429 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 430 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 5e-06, |
| "loss": 0.0113, |
| "step": 431 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 5e-06, |
| "loss": 0.0339, |
| "step": 432 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 5e-06, |
| "loss": 0.0099, |
| "step": 433 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0139, |
| "step": 434 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0099, |
| "step": 435 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0057, |
| "step": 436 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0066, |
| "step": 437 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0197, |
| "step": 438 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0205, |
| "step": 439 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0044, |
| "step": 440 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0253, |
| "step": 441 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0066, |
| "step": 442 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0107, |
| "step": 443 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 444 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 5e-06, |
| "loss": 0.0084, |
| "step": 445 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 5e-06, |
| "loss": 0.0353, |
| "step": 446 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 447 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0156, |
| "step": 448 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0135, |
| "step": 449 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0064, |
| "step": 450 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 451 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0034, |
| "step": 452 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 5e-06, |
| "loss": 0.02, |
| "step": 453 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 454 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 5e-06, |
| "loss": 0.0118, |
| "step": 455 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 456 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 5e-06, |
| "loss": 0.0146, |
| "step": 457 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 5e-06, |
| "loss": 0.0038, |
| "step": 458 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 5e-06, |
| "loss": 0.006, |
| "step": 459 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 5e-06, |
| "loss": 0.0076, |
| "step": 460 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 5e-06, |
| "loss": 0.0063, |
| "step": 461 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 5e-06, |
| "loss": 0.0211, |
| "step": 462 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 5e-06, |
| "loss": 0.024, |
| "step": 463 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 5e-06, |
| "loss": 0.0331, |
| "step": 464 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 5e-06, |
| "loss": 0.0239, |
| "step": 465 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 5e-06, |
| "loss": 0.0073, |
| "step": 466 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0063, |
| "step": 467 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 468 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0099, |
| "step": 469 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0071, |
| "step": 470 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0071, |
| "step": 471 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 472 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 5e-06, |
| "loss": 0.0066, |
| "step": 473 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 5e-06, |
| "loss": 0.0118, |
| "step": 474 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 5e-06, |
| "loss": 0.0306, |
| "step": 475 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 476 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 5e-06, |
| "loss": 0.0119, |
| "step": 477 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 5e-06, |
| "loss": 0.004, |
| "step": 478 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 5e-06, |
| "loss": 0.0149, |
| "step": 479 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 5e-06, |
| "loss": 0.0116, |
| "step": 480 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0125, |
| "step": 481 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0227, |
| "step": 482 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0275, |
| "step": 483 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0174, |
| "step": 484 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0117, |
| "step": 485 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0038, |
| "step": 486 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 5e-06, |
| "loss": 0.017, |
| "step": 487 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0155, |
| "step": 488 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0109, |
| "step": 489 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 490 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 5e-06, |
| "loss": 0.0196, |
| "step": 491 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 5e-06, |
| "loss": 0.0337, |
| "step": 492 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 5e-06, |
| "loss": 0.0211, |
| "step": 493 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 494 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 5e-06, |
| "loss": 0.0068, |
| "step": 495 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 5e-06, |
| "loss": 0.0215, |
| "step": 496 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 5e-06, |
| "loss": 0.011, |
| "step": 497 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 5e-06, |
| "loss": 0.0087, |
| "step": 498 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 5e-06, |
| "loss": 0.0251, |
| "step": 499 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0228, |
| "step": 500 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 501 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0041, |
| "step": 502 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0041, |
| "step": 503 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0113, |
| "step": 504 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 5e-06, |
| "loss": 0.0079, |
| "step": 505 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 5e-06, |
| "loss": 0.0069, |
| "step": 506 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 5e-06, |
| "loss": 0.0065, |
| "step": 507 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 5e-06, |
| "loss": 0.004, |
| "step": 508 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0261, |
| "step": 509 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 510 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0033, |
| "step": 511 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0045, |
| "step": 512 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0146, |
| "step": 513 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 514 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 515 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0167, |
| "step": 516 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0151, |
| "step": 517 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0154, |
| "step": 518 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0214, |
| "step": 519 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0095, |
| "step": 520 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0094, |
| "step": 521 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0159, |
| "step": 522 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0153, |
| "step": 523 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0141, |
| "step": 524 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0214, |
| "step": 525 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0173, |
| "step": 526 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0055, |
| "step": 527 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0149, |
| "step": 528 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0169, |
| "step": 529 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0138, |
| "step": 530 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 531 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0068, |
| "step": 532 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 533 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 5e-06, |
| "loss": 0.012, |
| "step": 534 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0073, |
| "step": 535 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0074, |
| "step": 536 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0141, |
| "step": 537 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 5e-06, |
| "loss": 0.015, |
| "step": 538 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0151, |
| "step": 539 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0112, |
| "step": 540 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0191, |
| "step": 541 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0015, |
| "step": 542 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0113, |
| "step": 543 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0024, |
| "step": 544 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0243, |
| "step": 545 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 546 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0088, |
| "step": 547 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 548 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0046, |
| "step": 549 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0088, |
| "step": 550 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 551 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 552 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0071, |
| "step": 553 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0152, |
| "step": 554 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0054, |
| "step": 555 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0016, |
| "step": 556 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0109, |
| "step": 557 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0129, |
| "step": 558 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 559 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0143, |
| "step": 560 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0221, |
| "step": 561 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0072, |
| "step": 562 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0074, |
| "step": 563 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0033, |
| "step": 564 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0146, |
| "step": 565 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0239, |
| "step": 566 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 567 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0226, |
| "step": 568 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0079, |
| "step": 569 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0294, |
| "step": 570 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0041, |
| "step": 571 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0252, |
| "step": 572 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0034, |
| "step": 573 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0133, |
| "step": 574 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0072, |
| "step": 575 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 5e-06, |
| "loss": 0.017, |
| "step": 576 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0125, |
| "step": 577 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0059, |
| "step": 578 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0062, |
| "step": 579 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0123, |
| "step": 580 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 581 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0277, |
| "step": 582 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0103, |
| "step": 583 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0148, |
| "step": 584 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0082, |
| "step": 585 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0064, |
| "step": 586 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0081, |
| "step": 587 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 588 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0147, |
| "step": 589 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0073, |
| "step": 590 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 591 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 5e-06, |
| "loss": 0.003, |
| "step": 592 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0124, |
| "step": 593 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 594 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0033, |
| "step": 595 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 596 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0177, |
| "step": 597 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0084, |
| "step": 598 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0044, |
| "step": 599 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0071, |
| "step": 600 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 601 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0078, |
| "step": 602 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 603 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0029, |
| "step": 604 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0136, |
| "step": 605 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 606 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 5e-06, |
| "loss": 0.015, |
| "step": 607 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0191, |
| "step": 608 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0098, |
| "step": 609 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0202, |
| "step": 610 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0034, |
| "step": 611 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 612 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 613 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0171, |
| "step": 614 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0078, |
| "step": 615 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0114, |
| "step": 616 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0088, |
| "step": 617 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0123, |
| "step": 618 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0238, |
| "step": 619 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0121, |
| "step": 620 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0083, |
| "step": 621 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0256, |
| "step": 622 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0078, |
| "step": 623 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0068, |
| "step": 624 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0068, |
| "step": 625 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0085, |
| "step": 626 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0037, |
| "step": 627 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0102, |
| "step": 628 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0042, |
| "step": 629 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0079, |
| "step": 630 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0132, |
| "step": 631 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 5e-06, |
| "loss": 0.016, |
| "step": 632 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0023, |
| "step": 633 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0229, |
| "step": 634 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 5e-06, |
| "loss": 0.003, |
| "step": 635 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0077, |
| "step": 636 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0204, |
| "step": 637 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0144, |
| "step": 638 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0007, |
| "step": 639 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 640 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0028, |
| "step": 641 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0149, |
| "step": 642 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0253, |
| "step": 643 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 644 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 645 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0066, |
| "step": 646 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0017, |
| "step": 647 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0058, |
| "step": 648 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 649 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0139, |
| "step": 650 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 651 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 652 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 653 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0042, |
| "step": 654 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0115, |
| "step": 655 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0252, |
| "step": 656 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 5e-06, |
| "loss": 0.004, |
| "step": 657 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0007, |
| "step": 658 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0199, |
| "step": 659 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 5e-06, |
| "loss": 0.015, |
| "step": 660 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 661 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 662 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0083, |
| "step": 663 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0019, |
| "step": 664 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0097, |
| "step": 665 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 5e-06, |
| "loss": 0.017, |
| "step": 666 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0038, |
| "step": 667 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0145, |
| "step": 668 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0056, |
| "step": 669 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 670 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 671 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 5e-06, |
| "loss": 0.011, |
| "step": 672 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 673 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 674 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0103, |
| "step": 675 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0055, |
| "step": 676 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 677 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 678 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0101, |
| "step": 679 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 680 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0083, |
| "step": 681 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 5e-06, |
| "loss": 0.003, |
| "step": 682 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0035, |
| "step": 683 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0136, |
| "step": 684 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0115, |
| "step": 685 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0111, |
| "step": 686 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0168, |
| "step": 687 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 688 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0137, |
| "step": 689 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0145, |
| "step": 690 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0137, |
| "step": 691 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0083, |
| "step": 692 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0222, |
| "step": 693 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0069, |
| "step": 694 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0029, |
| "step": 695 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0058, |
| "step": 696 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0135, |
| "step": 697 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0076, |
| "step": 698 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 699 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 5e-06, |
| "loss": 0.027, |
| "step": 700 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0019, |
| "step": 701 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0139, |
| "step": 702 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 5e-06, |
| "loss": 0.008, |
| "step": 703 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 704 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0081, |
| "step": 705 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0025, |
| "step": 706 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0157, |
| "step": 707 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0235, |
| "step": 708 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0154, |
| "step": 709 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 710 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 711 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0021, |
| "step": 712 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0033, |
| "step": 713 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0146, |
| "step": 714 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 715 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0057, |
| "step": 716 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0038, |
| "step": 717 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 5e-06, |
| "loss": 0.008, |
| "step": 718 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0174, |
| "step": 719 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 720 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0042, |
| "step": 721 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0244, |
| "step": 722 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 5e-06, |
| "loss": 0.004, |
| "step": 723 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0073, |
| "step": 724 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0029, |
| "step": 725 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 726 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0084, |
| "step": 727 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 728 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 729 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0097, |
| "step": 730 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0206, |
| "step": 731 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 732 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0187, |
| "step": 733 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0045, |
| "step": 734 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 735 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0131, |
| "step": 736 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0131, |
| "step": 737 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 738 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 5e-06, |
| "loss": 0.031, |
| "step": 739 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0152, |
| "step": 740 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 741 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0063, |
| "step": 742 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 743 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0127, |
| "step": 744 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0017, |
| "step": 745 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0058, |
| "step": 746 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0097, |
| "step": 747 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 5e-06, |
| "loss": 0.0015, |
| "step": 748 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 5e-06, |
| "loss": 0.0099, |
| "step": 749 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 5e-06, |
| "loss": 0.0138, |
| "step": 750 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 5e-06, |
| "loss": 0.0018, |
| "step": 751 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 5e-06, |
| "loss": 0.009, |
| "step": 752 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 5e-06, |
| "loss": 0.0107, |
| "step": 753 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 5e-06, |
| "loss": 0.0221, |
| "step": 754 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 755 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 5e-06, |
| "loss": 0.0269, |
| "step": 756 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 5e-06, |
| "loss": 0.0106, |
| "step": 757 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 5e-06, |
| "loss": 0.0114, |
| "step": 758 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 759 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 5e-06, |
| "loss": 0.011, |
| "step": 760 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 5e-06, |
| "loss": 0.0063, |
| "step": 761 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 762 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 763 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 5e-06, |
| "loss": 0.0089, |
| "step": 764 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 5e-06, |
| "loss": 0.0083, |
| "step": 765 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 5e-06, |
| "loss": 0.0025, |
| "step": 766 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 5e-06, |
| "loss": 0.0049, |
| "step": 767 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 5e-06, |
| "loss": 0.0006, |
| "step": 768 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 769 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 5e-06, |
| "loss": 0.005, |
| "step": 770 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 5e-06, |
| "loss": 0.0185, |
| "step": 771 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 772 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0082, |
| "step": 773 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0101, |
| "step": 774 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0147, |
| "step": 775 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 5e-06, |
| "loss": 0.0075, |
| "step": 776 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 5e-06, |
| "loss": 0.0178, |
| "step": 777 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 5e-06, |
| "loss": 0.003, |
| "step": 778 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 779 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 780 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 5e-06, |
| "loss": 0.009, |
| "step": 781 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 5e-06, |
| "loss": 0.014, |
| "step": 782 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 5e-06, |
| "loss": 0.0059, |
| "step": 783 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 5e-06, |
| "loss": 0.002, |
| "step": 784 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 785 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0182, |
| "step": 786 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0086, |
| "step": 787 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0072, |
| "step": 788 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0034, |
| "step": 789 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 790 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 5e-06, |
| "loss": 0.0093, |
| "step": 791 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 5e-06, |
| "loss": 0.0144, |
| "step": 792 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 5e-06, |
| "loss": 0.0112, |
| "step": 793 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 5e-06, |
| "loss": 0.0269, |
| "step": 794 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0028, |
| "step": 795 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0049, |
| "step": 796 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0101, |
| "step": 797 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0112, |
| "step": 798 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 5e-06, |
| "loss": 0.0074, |
| "step": 799 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0024, |
| "step": 800 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 801 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0017, |
| "step": 802 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0101, |
| "step": 803 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 5e-06, |
| "loss": 0.0095, |
| "step": 804 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0172, |
| "step": 805 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0078, |
| "step": 806 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0019, |
| "step": 807 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0147, |
| "step": 808 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 5e-06, |
| "loss": 0.0146, |
| "step": 809 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 810 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 811 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 5e-06, |
| "loss": 0.0131, |
| "step": 812 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 5e-06, |
| "loss": 0.0107, |
| "step": 813 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0092, |
| "step": 814 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 815 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 816 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 817 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 5e-06, |
| "loss": 0.0097, |
| "step": 818 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0064, |
| "step": 819 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 820 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0066, |
| "step": 821 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0129, |
| "step": 822 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 5e-06, |
| "loss": 0.0064, |
| "step": 823 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 5e-06, |
| "loss": 0.0109, |
| "step": 824 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 5e-06, |
| "loss": 0.0208, |
| "step": 825 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 5e-06, |
| "loss": 0.0019, |
| "step": 826 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 827 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 828 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 829 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0075, |
| "step": 830 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0105, |
| "step": 831 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 5e-06, |
| "loss": 0.0134, |
| "step": 832 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0081, |
| "step": 833 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 834 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 835 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 836 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 5e-06, |
| "loss": 0.0009, |
| "step": 837 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 5e-06, |
| "loss": 0.0088, |
| "step": 838 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 5e-06, |
| "loss": 0.0005, |
| "step": 839 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 840 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 5e-06, |
| "loss": 0.0033, |
| "step": 841 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0073, |
| "step": 842 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0167, |
| "step": 843 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0136, |
| "step": 844 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0048, |
| "step": 845 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 5e-06, |
| "loss": 0.0059, |
| "step": 846 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 847 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 5e-06, |
| "loss": 0.0004, |
| "step": 848 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 5e-06, |
| "loss": 0.013, |
| "step": 849 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 5e-06, |
| "loss": 0.0019, |
| "step": 850 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 851 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 852 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 5e-06, |
| "loss": 0.0138, |
| "step": 853 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 5e-06, |
| "loss": 0.0152, |
| "step": 854 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 5e-06, |
| "loss": 0.0093, |
| "step": 855 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 856 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0005, |
| "step": 857 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 5e-06, |
| "loss": 0.013, |
| "step": 858 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0016, |
| "step": 859 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 5e-06, |
| "loss": 0.0047, |
| "step": 860 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0054, |
| "step": 861 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0136, |
| "step": 862 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 5e-06, |
| "loss": 0.004, |
| "step": 863 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0199, |
| "step": 864 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 5e-06, |
| "loss": 0.0109, |
| "step": 865 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 5e-06, |
| "loss": 0.0091, |
| "step": 866 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 5e-06, |
| "loss": 0.0038, |
| "step": 867 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 5e-06, |
| "loss": 0.0064, |
| "step": 868 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 5e-06, |
| "loss": 0.003, |
| "step": 869 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 5e-06, |
| "loss": 0.0065, |
| "step": 870 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 5e-06, |
| "loss": 0.0038, |
| "step": 871 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 5e-06, |
| "loss": 0.0047, |
| "step": 872 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 5e-06, |
| "loss": 0.0108, |
| "step": 873 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 874 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 875 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 876 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 5e-06, |
| "loss": 0.0146, |
| "step": 877 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 878 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 5e-06, |
| "loss": 0.0174, |
| "step": 879 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 880 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 5e-06, |
| "loss": 0.0175, |
| "step": 881 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 5e-06, |
| "loss": 0.0006, |
| "step": 882 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 883 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 5e-06, |
| "loss": 0.0257, |
| "step": 884 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 5e-06, |
| "loss": 0.0052, |
| "step": 885 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 886 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 5e-06, |
| "loss": 0.0016, |
| "step": 887 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 5e-06, |
| "loss": 0.008, |
| "step": 888 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 5e-06, |
| "loss": 0.0077, |
| "step": 889 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 5e-06, |
| "loss": 0.0024, |
| "step": 890 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 5e-06, |
| "loss": 0.002, |
| "step": 891 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 5e-06, |
| "loss": 0.017, |
| "step": 892 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 5e-06, |
| "loss": 0.0083, |
| "step": 893 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 5e-06, |
| "loss": 0.0075, |
| "step": 894 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 5e-06, |
| "loss": 0.0055, |
| "step": 895 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 5e-06, |
| "loss": 0.0045, |
| "step": 896 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 5e-06, |
| "loss": 0.0047, |
| "step": 897 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 898 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 5e-06, |
| "loss": 0.0099, |
| "step": 899 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 5e-06, |
| "loss": 0.0111, |
| "step": 900 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 5e-06, |
| "loss": 0.002, |
| "step": 901 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 5e-06, |
| "loss": 0.0071, |
| "step": 902 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0023, |
| "step": 903 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0143, |
| "step": 904 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0106, |
| "step": 905 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5e-06, |
| "loss": 0.0113, |
| "step": 906 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5e-06, |
| "loss": 0.006, |
| "step": 907 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 908 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0021, |
| "step": 909 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0079, |
| "step": 910 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0151, |
| "step": 911 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 912 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 913 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5e-06, |
| "loss": 0.0008, |
| "step": 914 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5e-06, |
| "loss": 0.0133, |
| "step": 915 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5e-06, |
| "loss": 0.009, |
| "step": 916 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0015, |
| "step": 917 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0028, |
| "step": 918 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0014, |
| "step": 919 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0029, |
| "step": 920 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5e-06, |
| "loss": 0.0076, |
| "step": 921 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5e-06, |
| "loss": 0.0061, |
| "step": 922 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 923 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 924 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5e-06, |
| "loss": 0.015, |
| "step": 925 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5e-06, |
| "loss": 0.0007, |
| "step": 926 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5e-06, |
| "loss": 0.009, |
| "step": 927 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5e-06, |
| "loss": 0.0006, |
| "step": 928 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5e-06, |
| "loss": 0.0004, |
| "step": 929 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5e-06, |
| "loss": 0.0073, |
| "step": 930 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 931 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5e-06, |
| "loss": 0.0131, |
| "step": 932 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 933 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 934 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5e-06, |
| "loss": 0.0094, |
| "step": 935 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0009, |
| "step": 936 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0071, |
| "step": 937 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5e-06, |
| "loss": 0.018, |
| "step": 938 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0087, |
| "step": 939 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5e-06, |
| "loss": 0.0111, |
| "step": 940 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5e-06, |
| "loss": 0.007, |
| "step": 941 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 942 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5e-06, |
| "loss": 0.008, |
| "step": 943 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5e-06, |
| "loss": 0.0015, |
| "step": 944 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5e-06, |
| "loss": 0.0016, |
| "step": 945 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 946 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5e-06, |
| "loss": 0.0273, |
| "step": 947 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 948 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5e-06, |
| "loss": 0.0029, |
| "step": 949 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 950 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0016, |
| "step": 951 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5e-06, |
| "loss": 0.007, |
| "step": 952 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0153, |
| "step": 953 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5e-06, |
| "loss": 0.0017, |
| "step": 954 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 955 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0055, |
| "step": 956 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0014, |
| "step": 957 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 958 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 5e-06, |
| "loss": 0.0074, |
| "step": 959 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 5e-06, |
| "loss": 0.0064, |
| "step": 960 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 5e-06, |
| "loss": 0.0151, |
| "step": 961 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 962 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 5e-06, |
| "loss": 0.0064, |
| "step": 963 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 5e-06, |
| "loss": 0.007, |
| "step": 964 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 965 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 5e-06, |
| "loss": 0.0118, |
| "step": 966 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 5e-06, |
| "loss": 0.0087, |
| "step": 967 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 5e-06, |
| "loss": 0.0006, |
| "step": 968 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0016, |
| "step": 969 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0014, |
| "step": 970 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0016, |
| "step": 971 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 5e-06, |
| "loss": 0.008, |
| "step": 972 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 5e-06, |
| "loss": 0.0048, |
| "step": 973 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 974 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 5e-06, |
| "loss": 0.0092, |
| "step": 975 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 976 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 5e-06, |
| "loss": 0.0193, |
| "step": 977 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0133, |
| "step": 978 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0165, |
| "step": 979 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 980 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0017, |
| "step": 981 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 5e-06, |
| "loss": 0.0085, |
| "step": 982 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 983 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 984 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0024, |
| "step": 985 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0006, |
| "step": 986 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 987 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0004, |
| "step": 988 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 989 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0008, |
| "step": 990 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 991 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 992 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 993 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0091, |
| "step": 994 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0112, |
| "step": 995 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 5e-06, |
| "loss": 0.0025, |
| "step": 996 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 5e-06, |
| "loss": 0.004, |
| "step": 997 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0004, |
| "step": 998 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 999 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0057, |
| "step": 1000 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 5e-06, |
| "loss": 0.0034, |
| "step": 1001 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0035, |
| "step": 1002 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0003, |
| "step": 1003 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0189, |
| "step": 1004 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0028, |
| "step": 1005 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 5e-06, |
| "loss": 0.0017, |
| "step": 1006 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0335, |
| "step": 1007 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0177, |
| "step": 1008 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0007, |
| "step": 1009 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 5e-06, |
| "loss": 0.0072, |
| "step": 1010 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 1011 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 1012 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 1013 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 5e-06, |
| "loss": 0.0018, |
| "step": 1014 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 5e-06, |
| "loss": 0.002, |
| "step": 1015 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0037, |
| "step": 1016 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0153, |
| "step": 1017 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0091, |
| "step": 1018 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0115, |
| "step": 1019 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 1020 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 5e-06, |
| "loss": 0.003, |
| "step": 1021 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 5e-06, |
| "loss": 0.002, |
| "step": 1022 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 1023 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 5e-06, |
| "loss": 0.0107, |
| "step": 1024 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0096, |
| "step": 1025 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0098, |
| "step": 1026 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0151, |
| "step": 1027 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0113, |
| "step": 1028 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1029 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0086, |
| "step": 1030 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0065, |
| "step": 1031 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0106, |
| "step": 1032 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1033 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 1034 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 1035 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0084, |
| "step": 1036 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0099, |
| "step": 1037 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 1038 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0117, |
| "step": 1039 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0172, |
| "step": 1040 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0028, |
| "step": 1041 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0066, |
| "step": 1042 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 5e-06, |
| "loss": 0.0065, |
| "step": 1043 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 5e-06, |
| "loss": 0.004, |
| "step": 1044 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0089, |
| "step": 1045 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0114, |
| "step": 1046 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0098, |
| "step": 1047 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 1048 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0014, |
| "step": 1049 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0029, |
| "step": 1050 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0094, |
| "step": 1051 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 5e-06, |
| "loss": 0.0019, |
| "step": 1052 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0121, |
| "step": 1053 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 1054 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0008, |
| "step": 1055 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0059, |
| "step": 1056 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 5e-06, |
| "loss": 0.0095, |
| "step": 1057 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0003, |
| "step": 1058 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0045, |
| "step": 1059 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0029, |
| "step": 1060 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 1061 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 5e-06, |
| "loss": 0.0048, |
| "step": 1062 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 1063 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 5e-06, |
| "loss": 0.013, |
| "step": 1064 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0141, |
| "step": 1065 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 5e-06, |
| "loss": 0.0048, |
| "step": 1066 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0076, |
| "step": 1067 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 1068 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 5e-06, |
| "loss": 0.01, |
| "step": 1069 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0128, |
| "step": 1070 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 5e-06, |
| "loss": 0.0025, |
| "step": 1071 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0034, |
| "step": 1072 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0049, |
| "step": 1073 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0029, |
| "step": 1074 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0025, |
| "step": 1075 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 1076 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0125, |
| "step": 1077 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0008, |
| "step": 1078 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0121, |
| "step": 1079 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 5e-06, |
| "loss": 0.0052, |
| "step": 1080 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 1081 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 5e-06, |
| "loss": 0.009, |
| "step": 1082 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0076, |
| "step": 1083 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0007, |
| "step": 1084 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 1085 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0085, |
| "step": 1086 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0149, |
| "step": 1087 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0102, |
| "step": 1088 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0085, |
| "step": 1089 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 5e-06, |
| "loss": 0.0015, |
| "step": 1090 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0087, |
| "step": 1091 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0032, |
| "step": 1092 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1093 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0084, |
| "step": 1094 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 5e-06, |
| "loss": 0.0082, |
| "step": 1095 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0205, |
| "step": 1096 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0056, |
| "step": 1097 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 5e-06, |
| "loss": 0.014, |
| "step": 1098 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 5e-06, |
| "loss": 0.0141, |
| "step": 1099 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0044, |
| "step": 1100 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0142, |
| "step": 1101 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 1102 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0039, |
| "step": 1103 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 5e-06, |
| "loss": 0.0118, |
| "step": 1104 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0091, |
| "step": 1105 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 1106 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0052, |
| "step": 1107 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 1108 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 5e-06, |
| "loss": 0.0051, |
| "step": 1109 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0022, |
| "step": 1110 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0078, |
| "step": 1111 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0095, |
| "step": 1112 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 5e-06, |
| "loss": 0.0023, |
| "step": 1113 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0128, |
| "step": 1114 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0081, |
| "step": 1115 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0179, |
| "step": 1116 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0103, |
| "step": 1117 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 5e-06, |
| "loss": 0.0077, |
| "step": 1118 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0159, |
| "step": 1119 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0083, |
| "step": 1120 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0016, |
| "step": 1121 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0081, |
| "step": 1122 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 5e-06, |
| "loss": 0.0047, |
| "step": 1123 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0091, |
| "step": 1124 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 1125 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0036, |
| "step": 1126 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 5e-06, |
| "loss": 0.0017, |
| "step": 1127 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0136, |
| "step": 1128 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0052, |
| "step": 1129 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 1130 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0049, |
| "step": 1131 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 5e-06, |
| "loss": 0.0006, |
| "step": 1132 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1133 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0151, |
| "step": 1134 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 1135 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 5e-06, |
| "loss": 0.0141, |
| "step": 1136 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 1137 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0155, |
| "step": 1138 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0045, |
| "step": 1139 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1140 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0115, |
| "step": 1141 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 1142 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0018, |
| "step": 1143 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0021, |
| "step": 1144 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0026, |
| "step": 1145 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 5e-06, |
| "loss": 0.0028, |
| "step": 1146 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0031, |
| "step": 1147 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0166, |
| "step": 1148 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0021, |
| "step": 1149 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 5e-06, |
| "loss": 0.004, |
| "step": 1150 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 5e-06, |
| "loss": 0.0067, |
| "step": 1151 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0092, |
| "step": 1152 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0134, |
| "step": 1153 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0084, |
| "step": 1154 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0053, |
| "step": 1155 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 5e-06, |
| "loss": 0.0009, |
| "step": 1156 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1157 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0017, |
| "step": 1158 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0027, |
| "step": 1159 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 5e-06, |
| "loss": 0.0078, |
| "step": 1160 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0012, |
| "step": 1161 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0042, |
| "step": 1162 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0065, |
| "step": 1163 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0021, |
| "step": 1164 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1165 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0007, |
| "step": 1166 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0107, |
| "step": 1167 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0021, |
| "step": 1168 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0149, |
| "step": 1169 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 5e-06, |
| "loss": 0.0287, |
| "step": 1170 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1171 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0033, |
| "step": 1172 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0041, |
| "step": 1173 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1174 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0025, |
| "step": 1175 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0008, |
| "step": 1176 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0065, |
| "step": 1177 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0014, |
| "step": 1178 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 5e-06, |
| "loss": 0.0163, |
| "step": 1179 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 5e-06, |
| "loss": 0.001, |
| "step": 1180 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0025, |
| "step": 1181 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0085, |
| "step": 1182 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 1183 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 5e-06, |
| "loss": 0.0075, |
| "step": 1184 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0046, |
| "step": 1185 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0065, |
| "step": 1186 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0058, |
| "step": 1187 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 5e-06, |
| "loss": 0.0013, |
| "step": 1188 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0023, |
| "step": 1189 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0043, |
| "step": 1190 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0104, |
| "step": 1191 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0109, |
| "step": 1192 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 1193 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0007, |
| "step": 1194 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0009, |
| "step": 1195 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 5e-06, |
| "loss": 0.003, |
| "step": 1196 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0011, |
| "step": 1197 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 5e-06, |
| "loss": 0.0042, |
| "step": 1198 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0021, |
| "step": 1199 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 5e-06, |
| "loss": 0.0018, |
| "step": 1200 |
| } |
| ], |
| "logging_steps": 1.0, |
| "max_steps": 2345, |
| "num_train_epochs": 5, |
| "save_steps": 100, |
| "total_flos": 1.223272227101737e+19, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|