| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 40.0, |
| "global_step": 1000, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.04, |
| "learning_rate": 2.0000000000000003e-06, |
| "loss": 2.6464, |
| "step": 1 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 4.000000000000001e-06, |
| "loss": 2.6979, |
| "step": 2 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 6e-06, |
| "loss": 3.0822, |
| "step": 3 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 8.000000000000001e-06, |
| "loss": 2.7994, |
| "step": 4 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 1e-05, |
| "loss": 2.8679, |
| "step": 5 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 1.2e-05, |
| "loss": 2.5602, |
| "step": 6 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 1.4000000000000001e-05, |
| "loss": 2.9689, |
| "step": 7 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 1.6000000000000003e-05, |
| "loss": 3.068, |
| "step": 8 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 1.8e-05, |
| "loss": 2.6173, |
| "step": 9 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 2e-05, |
| "loss": 2.8631, |
| "step": 10 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 2.2000000000000003e-05, |
| "loss": 2.7556, |
| "step": 11 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 2.4e-05, |
| "loss": 2.8262, |
| "step": 12 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 2.6000000000000002e-05, |
| "loss": 2.8098, |
| "step": 13 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 2.8000000000000003e-05, |
| "loss": 3.0013, |
| "step": 14 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 3e-05, |
| "loss": 2.8608, |
| "step": 15 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 3.2000000000000005e-05, |
| "loss": 2.7332, |
| "step": 16 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 3.4000000000000007e-05, |
| "loss": 2.4835, |
| "step": 17 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 3.6e-05, |
| "loss": 3.0711, |
| "step": 18 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 3.8e-05, |
| "loss": 2.7769, |
| "step": 19 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 4e-05, |
| "loss": 2.7375, |
| "step": 20 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 4.2e-05, |
| "loss": 2.6312, |
| "step": 21 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 4.4000000000000006e-05, |
| "loss": 2.4671, |
| "step": 22 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 4.600000000000001e-05, |
| "loss": 2.6448, |
| "step": 23 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 4.8e-05, |
| "loss": 2.7383, |
| "step": 24 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 5e-05, |
| "loss": 2.8138, |
| "step": 25 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 5.2000000000000004e-05, |
| "loss": 2.7637, |
| "step": 26 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 5.4000000000000005e-05, |
| "loss": 2.8317, |
| "step": 27 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 5.6000000000000006e-05, |
| "loss": 2.6637, |
| "step": 28 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 5.8e-05, |
| "loss": 2.5164, |
| "step": 29 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 6e-05, |
| "loss": 2.6832, |
| "step": 30 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 6.2e-05, |
| "loss": 2.3638, |
| "step": 31 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 6.400000000000001e-05, |
| "loss": 2.4183, |
| "step": 32 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 6.6e-05, |
| "loss": 2.6921, |
| "step": 33 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 6.800000000000001e-05, |
| "loss": 2.3107, |
| "step": 34 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 7e-05, |
| "loss": 2.5473, |
| "step": 35 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 7.2e-05, |
| "loss": 2.3215, |
| "step": 36 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 7.4e-05, |
| "loss": 2.3483, |
| "step": 37 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 7.6e-05, |
| "loss": 2.3435, |
| "step": 38 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 7.800000000000001e-05, |
| "loss": 2.2673, |
| "step": 39 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 8e-05, |
| "loss": 2.3928, |
| "step": 40 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 8.2e-05, |
| "loss": 2.3331, |
| "step": 41 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 8.4e-05, |
| "loss": 2.2442, |
| "step": 42 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 8.6e-05, |
| "loss": 2.2282, |
| "step": 43 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 8.800000000000001e-05, |
| "loss": 2.4065, |
| "step": 44 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 9e-05, |
| "loss": 2.0132, |
| "step": 45 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 9.200000000000001e-05, |
| "loss": 2.0322, |
| "step": 46 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 9.4e-05, |
| "loss": 2.1801, |
| "step": 47 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 9.6e-05, |
| "loss": 1.8874, |
| "step": 48 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 9.8e-05, |
| "loss": 1.7562, |
| "step": 49 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 0.0001, |
| "loss": 1.7259, |
| "step": 50 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 0.00010200000000000001, |
| "loss": 1.7008, |
| "step": 51 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 0.00010400000000000001, |
| "loss": 1.5573, |
| "step": 52 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 0.00010600000000000002, |
| "loss": 1.7009, |
| "step": 53 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 0.00010800000000000001, |
| "loss": 1.6904, |
| "step": 54 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 0.00011000000000000002, |
| "loss": 1.6967, |
| "step": 55 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 0.00011200000000000001, |
| "loss": 1.5618, |
| "step": 56 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 0.00011399999999999999, |
| "loss": 1.3273, |
| "step": 57 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 0.000116, |
| "loss": 1.7243, |
| "step": 58 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 0.000118, |
| "loss": 1.6316, |
| "step": 59 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 0.00012, |
| "loss": 1.5554, |
| "step": 60 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 0.000122, |
| "loss": 1.5982, |
| "step": 61 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 0.000124, |
| "loss": 1.3389, |
| "step": 62 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 0.000126, |
| "loss": 1.5836, |
| "step": 63 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 0.00012800000000000002, |
| "loss": 1.3101, |
| "step": 64 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 0.00013000000000000002, |
| "loss": 1.3613, |
| "step": 65 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 0.000132, |
| "loss": 1.5751, |
| "step": 66 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 0.000134, |
| "loss": 1.2825, |
| "step": 67 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 0.00013600000000000003, |
| "loss": 1.6813, |
| "step": 68 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 0.000138, |
| "loss": 1.3686, |
| "step": 69 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 0.00014, |
| "loss": 1.5897, |
| "step": 70 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 0.000142, |
| "loss": 1.3348, |
| "step": 71 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 0.000144, |
| "loss": 1.6012, |
| "step": 72 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 0.000146, |
| "loss": 1.3558, |
| "step": 73 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 0.000148, |
| "loss": 1.6032, |
| "step": 74 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 0.00015000000000000001, |
| "loss": 1.3278, |
| "step": 75 |
| }, |
| { |
| "epoch": 3.04, |
| "learning_rate": 0.000152, |
| "loss": 1.0751, |
| "step": 76 |
| }, |
| { |
| "epoch": 3.08, |
| "learning_rate": 0.000154, |
| "loss": 1.1133, |
| "step": 77 |
| }, |
| { |
| "epoch": 3.12, |
| "learning_rate": 0.00015600000000000002, |
| "loss": 1.2006, |
| "step": 78 |
| }, |
| { |
| "epoch": 3.16, |
| "learning_rate": 0.00015800000000000002, |
| "loss": 1.4573, |
| "step": 79 |
| }, |
| { |
| "epoch": 3.2, |
| "learning_rate": 0.00016, |
| "loss": 1.3563, |
| "step": 80 |
| }, |
| { |
| "epoch": 3.24, |
| "learning_rate": 0.000162, |
| "loss": 1.4274, |
| "step": 81 |
| }, |
| { |
| "epoch": 3.28, |
| "learning_rate": 0.000164, |
| "loss": 1.1248, |
| "step": 82 |
| }, |
| { |
| "epoch": 3.32, |
| "learning_rate": 0.000166, |
| "loss": 0.9875, |
| "step": 83 |
| }, |
| { |
| "epoch": 3.36, |
| "learning_rate": 0.000168, |
| "loss": 1.1229, |
| "step": 84 |
| }, |
| { |
| "epoch": 3.4, |
| "learning_rate": 0.00017, |
| "loss": 1.1158, |
| "step": 85 |
| }, |
| { |
| "epoch": 3.44, |
| "learning_rate": 0.000172, |
| "loss": 1.1138, |
| "step": 86 |
| }, |
| { |
| "epoch": 3.48, |
| "learning_rate": 0.000174, |
| "loss": 1.218, |
| "step": 87 |
| }, |
| { |
| "epoch": 3.52, |
| "learning_rate": 0.00017600000000000002, |
| "loss": 1.2347, |
| "step": 88 |
| }, |
| { |
| "epoch": 3.56, |
| "learning_rate": 0.00017800000000000002, |
| "loss": 1.1635, |
| "step": 89 |
| }, |
| { |
| "epoch": 3.6, |
| "learning_rate": 0.00018, |
| "loss": 1.1109, |
| "step": 90 |
| }, |
| { |
| "epoch": 3.64, |
| "learning_rate": 0.000182, |
| "loss": 1.2611, |
| "step": 91 |
| }, |
| { |
| "epoch": 3.68, |
| "learning_rate": 0.00018400000000000003, |
| "loss": 1.1428, |
| "step": 92 |
| }, |
| { |
| "epoch": 3.72, |
| "learning_rate": 0.00018600000000000002, |
| "loss": 1.0715, |
| "step": 93 |
| }, |
| { |
| "epoch": 3.76, |
| "learning_rate": 0.000188, |
| "loss": 0.8545, |
| "step": 94 |
| }, |
| { |
| "epoch": 3.8, |
| "learning_rate": 0.00019, |
| "loss": 0.9033, |
| "step": 95 |
| }, |
| { |
| "epoch": 3.84, |
| "learning_rate": 0.000192, |
| "loss": 0.9245, |
| "step": 96 |
| }, |
| { |
| "epoch": 3.88, |
| "learning_rate": 0.000194, |
| "loss": 0.8899, |
| "step": 97 |
| }, |
| { |
| "epoch": 3.92, |
| "learning_rate": 0.000196, |
| "loss": 0.9467, |
| "step": 98 |
| }, |
| { |
| "epoch": 3.96, |
| "learning_rate": 0.00019800000000000002, |
| "loss": 0.8061, |
| "step": 99 |
| }, |
| { |
| "epoch": 4.0, |
| "learning_rate": 0.0002, |
| "loss": 1.0404, |
| "step": 100 |
| }, |
| { |
| "epoch": 4.04, |
| "learning_rate": 0.00019999986330190926, |
| "loss": 0.8937, |
| "step": 101 |
| }, |
| { |
| "epoch": 4.08, |
| "learning_rate": 0.00019999945320801072, |
| "loss": 0.9378, |
| "step": 102 |
| }, |
| { |
| "epoch": 4.12, |
| "learning_rate": 0.00019999876971942557, |
| "loss": 0.7405, |
| "step": 103 |
| }, |
| { |
| "epoch": 4.16, |
| "learning_rate": 0.0001999978128380225, |
| "loss": 0.6425, |
| "step": 104 |
| }, |
| { |
| "epoch": 4.2, |
| "learning_rate": 0.00019999658256641747, |
| "loss": 0.7256, |
| "step": 105 |
| }, |
| { |
| "epoch": 4.24, |
| "learning_rate": 0.00019999507890797408, |
| "loss": 0.7782, |
| "step": 106 |
| }, |
| { |
| "epoch": 4.28, |
| "learning_rate": 0.0001999933018668033, |
| "loss": 0.5233, |
| "step": 107 |
| }, |
| { |
| "epoch": 4.32, |
| "learning_rate": 0.0001999912514477634, |
| "loss": 0.5725, |
| "step": 108 |
| }, |
| { |
| "epoch": 4.36, |
| "learning_rate": 0.00019998892765646026, |
| "loss": 0.7349, |
| "step": 109 |
| }, |
| { |
| "epoch": 4.4, |
| "learning_rate": 0.0001999863304992469, |
| "loss": 0.8363, |
| "step": 110 |
| }, |
| { |
| "epoch": 4.44, |
| "learning_rate": 0.00019998345998322397, |
| "loss": 0.6284, |
| "step": 111 |
| }, |
| { |
| "epoch": 4.48, |
| "learning_rate": 0.0001999803161162393, |
| "loss": 0.621, |
| "step": 112 |
| }, |
| { |
| "epoch": 4.52, |
| "learning_rate": 0.0001999768989068881, |
| "loss": 0.5174, |
| "step": 113 |
| }, |
| { |
| "epoch": 4.56, |
| "learning_rate": 0.0001999732083645129, |
| "loss": 0.4806, |
| "step": 114 |
| }, |
| { |
| "epoch": 4.6, |
| "learning_rate": 0.0001999692444992035, |
| "loss": 0.62, |
| "step": 115 |
| }, |
| { |
| "epoch": 4.64, |
| "learning_rate": 0.00019996500732179695, |
| "loss": 0.5301, |
| "step": 116 |
| }, |
| { |
| "epoch": 4.68, |
| "learning_rate": 0.0001999604968438775, |
| "loss": 0.5896, |
| "step": 117 |
| }, |
| { |
| "epoch": 4.72, |
| "learning_rate": 0.0001999557130777767, |
| "loss": 0.6004, |
| "step": 118 |
| }, |
| { |
| "epoch": 4.76, |
| "learning_rate": 0.00019995065603657316, |
| "loss": 0.4911, |
| "step": 119 |
| }, |
| { |
| "epoch": 4.8, |
| "learning_rate": 0.00019994532573409262, |
| "loss": 0.43, |
| "step": 120 |
| }, |
| { |
| "epoch": 4.84, |
| "learning_rate": 0.0001999397221849079, |
| "loss": 0.4462, |
| "step": 121 |
| }, |
| { |
| "epoch": 4.88, |
| "learning_rate": 0.00019993384540433894, |
| "loss": 0.4302, |
| "step": 122 |
| }, |
| { |
| "epoch": 4.92, |
| "learning_rate": 0.00019992769540845258, |
| "loss": 0.513, |
| "step": 123 |
| }, |
| { |
| "epoch": 4.96, |
| "learning_rate": 0.00019992127221406275, |
| "loss": 0.3951, |
| "step": 124 |
| }, |
| { |
| "epoch": 5.0, |
| "learning_rate": 0.0001999145758387301, |
| "loss": 0.4115, |
| "step": 125 |
| }, |
| { |
| "epoch": 5.04, |
| "learning_rate": 0.00019990760630076237, |
| "loss": 0.3703, |
| "step": 126 |
| }, |
| { |
| "epoch": 5.08, |
| "learning_rate": 0.000199900363619214, |
| "loss": 0.3361, |
| "step": 127 |
| }, |
| { |
| "epoch": 5.12, |
| "learning_rate": 0.00019989284781388617, |
| "loss": 0.2987, |
| "step": 128 |
| }, |
| { |
| "epoch": 5.16, |
| "learning_rate": 0.0001998850589053268, |
| "loss": 0.3532, |
| "step": 129 |
| }, |
| { |
| "epoch": 5.2, |
| "learning_rate": 0.00019987699691483048, |
| "loss": 0.2674, |
| "step": 130 |
| }, |
| { |
| "epoch": 5.24, |
| "learning_rate": 0.0001998686618644384, |
| "loss": 0.3665, |
| "step": 131 |
| }, |
| { |
| "epoch": 5.28, |
| "learning_rate": 0.00019986005377693825, |
| "loss": 0.297, |
| "step": 132 |
| }, |
| { |
| "epoch": 5.32, |
| "learning_rate": 0.00019985117267586424, |
| "loss": 0.3034, |
| "step": 133 |
| }, |
| { |
| "epoch": 5.36, |
| "learning_rate": 0.00019984201858549693, |
| "loss": 0.3665, |
| "step": 134 |
| }, |
| { |
| "epoch": 5.4, |
| "learning_rate": 0.00019983259153086327, |
| "loss": 0.2611, |
| "step": 135 |
| }, |
| { |
| "epoch": 5.44, |
| "learning_rate": 0.00019982289153773646, |
| "loss": 0.3011, |
| "step": 136 |
| }, |
| { |
| "epoch": 5.48, |
| "learning_rate": 0.00019981291863263592, |
| "loss": 0.3124, |
| "step": 137 |
| }, |
| { |
| "epoch": 5.52, |
| "learning_rate": 0.00019980267284282717, |
| "loss": 0.411, |
| "step": 138 |
| }, |
| { |
| "epoch": 5.56, |
| "learning_rate": 0.00019979215419632182, |
| "loss": 0.2738, |
| "step": 139 |
| }, |
| { |
| "epoch": 5.6, |
| "learning_rate": 0.00019978136272187747, |
| "loss": 0.236, |
| "step": 140 |
| }, |
| { |
| "epoch": 5.64, |
| "learning_rate": 0.00019977029844899758, |
| "loss": 0.28, |
| "step": 141 |
| }, |
| { |
| "epoch": 5.68, |
| "learning_rate": 0.00019975896140793142, |
| "loss": 0.2735, |
| "step": 142 |
| }, |
| { |
| "epoch": 5.72, |
| "learning_rate": 0.0001997473516296741, |
| "loss": 0.2348, |
| "step": 143 |
| }, |
| { |
| "epoch": 5.76, |
| "learning_rate": 0.00019973546914596623, |
| "loss": 0.2769, |
| "step": 144 |
| }, |
| { |
| "epoch": 5.8, |
| "learning_rate": 0.0001997233139892941, |
| "loss": 0.2491, |
| "step": 145 |
| }, |
| { |
| "epoch": 5.84, |
| "learning_rate": 0.0001997108861928895, |
| "loss": 0.2366, |
| "step": 146 |
| }, |
| { |
| "epoch": 5.88, |
| "learning_rate": 0.00019969818579072945, |
| "loss": 0.2332, |
| "step": 147 |
| }, |
| { |
| "epoch": 5.92, |
| "learning_rate": 0.00019968521281753642, |
| "loss": 0.3085, |
| "step": 148 |
| }, |
| { |
| "epoch": 5.96, |
| "learning_rate": 0.00019967196730877803, |
| "loss": 0.2387, |
| "step": 149 |
| }, |
| { |
| "epoch": 6.0, |
| "learning_rate": 0.000199658449300667, |
| "loss": 0.2472, |
| "step": 150 |
| }, |
| { |
| "epoch": 6.04, |
| "learning_rate": 0.000199644658830161, |
| "loss": 0.1949, |
| "step": 151 |
| }, |
| { |
| "epoch": 6.08, |
| "learning_rate": 0.00019963059593496268, |
| "loss": 0.1789, |
| "step": 152 |
| }, |
| { |
| "epoch": 6.12, |
| "learning_rate": 0.00019961626065351947, |
| "loss": 0.2397, |
| "step": 153 |
| }, |
| { |
| "epoch": 6.16, |
| "learning_rate": 0.0001996016530250235, |
| "loss": 0.2253, |
| "step": 154 |
| }, |
| { |
| "epoch": 6.2, |
| "learning_rate": 0.00019958677308941139, |
| "loss": 0.1792, |
| "step": 155 |
| }, |
| { |
| "epoch": 6.24, |
| "learning_rate": 0.0001995716208873644, |
| "loss": 0.1755, |
| "step": 156 |
| }, |
| { |
| "epoch": 6.28, |
| "learning_rate": 0.00019955619646030802, |
| "loss": 0.1779, |
| "step": 157 |
| }, |
| { |
| "epoch": 6.32, |
| "learning_rate": 0.00019954049985041204, |
| "loss": 0.2029, |
| "step": 158 |
| }, |
| { |
| "epoch": 6.36, |
| "learning_rate": 0.00019952453110059045, |
| "loss": 0.1681, |
| "step": 159 |
| }, |
| { |
| "epoch": 6.4, |
| "learning_rate": 0.00019950829025450114, |
| "loss": 0.1809, |
| "step": 160 |
| }, |
| { |
| "epoch": 6.44, |
| "learning_rate": 0.00019949177735654601, |
| "loss": 0.2218, |
| "step": 161 |
| }, |
| { |
| "epoch": 6.48, |
| "learning_rate": 0.00019947499245187068, |
| "loss": 0.174, |
| "step": 162 |
| }, |
| { |
| "epoch": 6.52, |
| "learning_rate": 0.00019945793558636437, |
| "loss": 0.2311, |
| "step": 163 |
| }, |
| { |
| "epoch": 6.56, |
| "learning_rate": 0.00019944060680666002, |
| "loss": 0.2011, |
| "step": 164 |
| }, |
| { |
| "epoch": 6.6, |
| "learning_rate": 0.0001994230061601338, |
| "loss": 0.2396, |
| "step": 165 |
| }, |
| { |
| "epoch": 6.64, |
| "learning_rate": 0.00019940513369490516, |
| "loss": 0.2079, |
| "step": 166 |
| }, |
| { |
| "epoch": 6.68, |
| "learning_rate": 0.00019938698945983676, |
| "loss": 0.1857, |
| "step": 167 |
| }, |
| { |
| "epoch": 6.72, |
| "learning_rate": 0.0001993685735045343, |
| "loss": 0.1772, |
| "step": 168 |
| }, |
| { |
| "epoch": 6.76, |
| "learning_rate": 0.00019934988587934623, |
| "loss": 0.1661, |
| "step": 169 |
| }, |
| { |
| "epoch": 6.8, |
| "learning_rate": 0.00019933092663536382, |
| "loss": 0.2161, |
| "step": 170 |
| }, |
| { |
| "epoch": 6.84, |
| "learning_rate": 0.00019931169582442095, |
| "loss": 0.2474, |
| "step": 171 |
| }, |
| { |
| "epoch": 6.88, |
| "learning_rate": 0.00019929219349909392, |
| "loss": 0.2145, |
| "step": 172 |
| }, |
| { |
| "epoch": 6.92, |
| "learning_rate": 0.0001992724197127013, |
| "loss": 0.2441, |
| "step": 173 |
| }, |
| { |
| "epoch": 6.96, |
| "learning_rate": 0.0001992523745193039, |
| "loss": 0.218, |
| "step": 174 |
| }, |
| { |
| "epoch": 7.0, |
| "learning_rate": 0.0001992320579737045, |
| "loss": 0.2064, |
| "step": 175 |
| }, |
| { |
| "epoch": 7.04, |
| "learning_rate": 0.0001992114701314478, |
| "loss": 0.1347, |
| "step": 176 |
| }, |
| { |
| "epoch": 7.08, |
| "learning_rate": 0.0001991906110488201, |
| "loss": 0.1503, |
| "step": 177 |
| }, |
| { |
| "epoch": 7.12, |
| "learning_rate": 0.0001991694807828494, |
| "loss": 0.1619, |
| "step": 178 |
| }, |
| { |
| "epoch": 7.16, |
| "learning_rate": 0.000199148079391305, |
| "loss": 0.1903, |
| "step": 179 |
| }, |
| { |
| "epoch": 7.2, |
| "learning_rate": 0.00019912640693269752, |
| "loss": 0.1756, |
| "step": 180 |
| }, |
| { |
| "epoch": 7.24, |
| "learning_rate": 0.00019910446346627862, |
| "loss": 0.1456, |
| "step": 181 |
| }, |
| { |
| "epoch": 7.28, |
| "learning_rate": 0.0001990822490520409, |
| "loss": 0.1534, |
| "step": 182 |
| }, |
| { |
| "epoch": 7.32, |
| "learning_rate": 0.00019905976375071772, |
| "loss": 0.1842, |
| "step": 183 |
| }, |
| { |
| "epoch": 7.36, |
| "learning_rate": 0.000199037007623783, |
| "loss": 0.1722, |
| "step": 184 |
| }, |
| { |
| "epoch": 7.4, |
| "learning_rate": 0.00019901398073345118, |
| "loss": 0.1718, |
| "step": 185 |
| }, |
| { |
| "epoch": 7.44, |
| "learning_rate": 0.00019899068314267688, |
| "loss": 0.2055, |
| "step": 186 |
| }, |
| { |
| "epoch": 7.48, |
| "learning_rate": 0.00019896711491515482, |
| "loss": 0.1626, |
| "step": 187 |
| }, |
| { |
| "epoch": 7.52, |
| "learning_rate": 0.0001989432761153196, |
| "loss": 0.158, |
| "step": 188 |
| }, |
| { |
| "epoch": 7.56, |
| "learning_rate": 0.00019891916680834566, |
| "loss": 0.2145, |
| "step": 189 |
| }, |
| { |
| "epoch": 7.6, |
| "learning_rate": 0.00019889478706014687, |
| "loss": 0.1673, |
| "step": 190 |
| }, |
| { |
| "epoch": 7.64, |
| "learning_rate": 0.00019887013693737653, |
| "loss": 0.1944, |
| "step": 191 |
| }, |
| { |
| "epoch": 7.68, |
| "learning_rate": 0.00019884521650742715, |
| "loss": 0.151, |
| "step": 192 |
| }, |
| { |
| "epoch": 7.72, |
| "learning_rate": 0.00019882002583843024, |
| "loss": 0.2067, |
| "step": 193 |
| }, |
| { |
| "epoch": 7.76, |
| "learning_rate": 0.00019879456499925614, |
| "loss": 0.1723, |
| "step": 194 |
| }, |
| { |
| "epoch": 7.8, |
| "learning_rate": 0.00019876883405951377, |
| "loss": 0.2001, |
| "step": 195 |
| }, |
| { |
| "epoch": 7.84, |
| "learning_rate": 0.00019874283308955057, |
| "loss": 0.1601, |
| "step": 196 |
| }, |
| { |
| "epoch": 7.88, |
| "learning_rate": 0.0001987165621604522, |
| "loss": 0.1882, |
| "step": 197 |
| }, |
| { |
| "epoch": 7.92, |
| "learning_rate": 0.00019869002134404235, |
| "loss": 0.1312, |
| "step": 198 |
| }, |
| { |
| "epoch": 7.96, |
| "learning_rate": 0.0001986632107128826, |
| "loss": 0.1808, |
| "step": 199 |
| }, |
| { |
| "epoch": 8.0, |
| "learning_rate": 0.00019863613034027224, |
| "loss": 0.1885, |
| "step": 200 |
| }, |
| { |
| "epoch": 8.04, |
| "learning_rate": 0.0001986087803002479, |
| "loss": 0.1291, |
| "step": 201 |
| }, |
| { |
| "epoch": 8.08, |
| "learning_rate": 0.00019858116066758362, |
| "loss": 0.1355, |
| "step": 202 |
| }, |
| { |
| "epoch": 8.12, |
| "learning_rate": 0.00019855327151779042, |
| "loss": 0.1746, |
| "step": 203 |
| }, |
| { |
| "epoch": 8.16, |
| "learning_rate": 0.00019852511292711608, |
| "loss": 0.1232, |
| "step": 204 |
| }, |
| { |
| "epoch": 8.2, |
| "learning_rate": 0.0001984966849725452, |
| "loss": 0.1544, |
| "step": 205 |
| }, |
| { |
| "epoch": 8.24, |
| "learning_rate": 0.00019846798773179866, |
| "loss": 0.1374, |
| "step": 206 |
| }, |
| { |
| "epoch": 8.28, |
| "learning_rate": 0.00019843902128333367, |
| "loss": 0.1781, |
| "step": 207 |
| }, |
| { |
| "epoch": 8.32, |
| "learning_rate": 0.0001984097857063434, |
| "loss": 0.1427, |
| "step": 208 |
| }, |
| { |
| "epoch": 8.36, |
| "learning_rate": 0.00019838028108075671, |
| "loss": 0.1684, |
| "step": 209 |
| }, |
| { |
| "epoch": 8.4, |
| "learning_rate": 0.00019835050748723824, |
| "loss": 0.1568, |
| "step": 210 |
| }, |
| { |
| "epoch": 8.44, |
| "learning_rate": 0.0001983204650071878, |
| "loss": 0.1202, |
| "step": 211 |
| }, |
| { |
| "epoch": 8.48, |
| "learning_rate": 0.00019829015372274038, |
| "loss": 0.1564, |
| "step": 212 |
| }, |
| { |
| "epoch": 8.52, |
| "learning_rate": 0.00019825957371676588, |
| "loss": 0.1683, |
| "step": 213 |
| }, |
| { |
| "epoch": 8.56, |
| "learning_rate": 0.0001982287250728689, |
| "loss": 0.1121, |
| "step": 214 |
| }, |
| { |
| "epoch": 8.6, |
| "learning_rate": 0.0001981976078753884, |
| "loss": 0.11, |
| "step": 215 |
| }, |
| { |
| "epoch": 8.64, |
| "learning_rate": 0.0001981662222093976, |
| "loss": 0.1358, |
| "step": 216 |
| }, |
| { |
| "epoch": 8.68, |
| "learning_rate": 0.0001981345681607038, |
| "loss": 0.2219, |
| "step": 217 |
| }, |
| { |
| "epoch": 8.72, |
| "learning_rate": 0.00019810264581584787, |
| "loss": 0.1534, |
| "step": 218 |
| }, |
| { |
| "epoch": 8.76, |
| "learning_rate": 0.00019807045526210436, |
| "loss": 0.1555, |
| "step": 219 |
| }, |
| { |
| "epoch": 8.8, |
| "learning_rate": 0.00019803799658748094, |
| "loss": 0.1126, |
| "step": 220 |
| }, |
| { |
| "epoch": 8.84, |
| "learning_rate": 0.00019800526988071843, |
| "loss": 0.1676, |
| "step": 221 |
| }, |
| { |
| "epoch": 8.88, |
| "learning_rate": 0.0001979722752312904, |
| "loss": 0.2178, |
| "step": 222 |
| }, |
| { |
| "epoch": 8.92, |
| "learning_rate": 0.00019793901272940293, |
| "loss": 0.188, |
| "step": 223 |
| }, |
| { |
| "epoch": 8.96, |
| "learning_rate": 0.00019790548246599447, |
| "loss": 0.2533, |
| "step": 224 |
| }, |
| { |
| "epoch": 9.0, |
| "learning_rate": 0.00019787168453273544, |
| "loss": 0.1704, |
| "step": 225 |
| }, |
| { |
| "epoch": 9.04, |
| "learning_rate": 0.00019783761902202813, |
| "loss": 0.1404, |
| "step": 226 |
| }, |
| { |
| "epoch": 9.08, |
| "learning_rate": 0.00019780328602700636, |
| "loss": 0.1118, |
| "step": 227 |
| }, |
| { |
| "epoch": 9.12, |
| "learning_rate": 0.00019776868564153516, |
| "loss": 0.1545, |
| "step": 228 |
| }, |
| { |
| "epoch": 9.16, |
| "learning_rate": 0.0001977338179602107, |
| "loss": 0.1021, |
| "step": 229 |
| }, |
| { |
| "epoch": 9.2, |
| "learning_rate": 0.00019769868307835994, |
| "loss": 0.1153, |
| "step": 230 |
| }, |
| { |
| "epoch": 9.24, |
| "learning_rate": 0.00019766328109204024, |
| "loss": 0.1167, |
| "step": 231 |
| }, |
| { |
| "epoch": 9.28, |
| "learning_rate": 0.00019762761209803927, |
| "loss": 0.136, |
| "step": 232 |
| }, |
| { |
| "epoch": 9.32, |
| "learning_rate": 0.00019759167619387476, |
| "loss": 0.1334, |
| "step": 233 |
| }, |
| { |
| "epoch": 9.36, |
| "learning_rate": 0.00019755547347779403, |
| "loss": 0.1369, |
| "step": 234 |
| }, |
| { |
| "epoch": 9.4, |
| "learning_rate": 0.000197519004048774, |
| "loss": 0.1274, |
| "step": 235 |
| }, |
| { |
| "epoch": 9.44, |
| "learning_rate": 0.0001974822680065206, |
| "loss": 0.124, |
| "step": 236 |
| }, |
| { |
| "epoch": 9.48, |
| "learning_rate": 0.00019744526545146886, |
| "loss": 0.1217, |
| "step": 237 |
| }, |
| { |
| "epoch": 9.52, |
| "learning_rate": 0.00019740799648478233, |
| "loss": 0.1376, |
| "step": 238 |
| }, |
| { |
| "epoch": 9.56, |
| "learning_rate": 0.0001973704612083529, |
| "loss": 0.1612, |
| "step": 239 |
| }, |
| { |
| "epoch": 9.6, |
| "learning_rate": 0.0001973326597248006, |
| "loss": 0.1675, |
| "step": 240 |
| }, |
| { |
| "epoch": 9.64, |
| "learning_rate": 0.00019729459213747326, |
| "loss": 0.1515, |
| "step": 241 |
| }, |
| { |
| "epoch": 9.68, |
| "learning_rate": 0.00019725625855044617, |
| "loss": 0.1457, |
| "step": 242 |
| }, |
| { |
| "epoch": 9.72, |
| "learning_rate": 0.00019721765906852197, |
| "loss": 0.2122, |
| "step": 243 |
| }, |
| { |
| "epoch": 9.76, |
| "learning_rate": 0.00019717879379723012, |
| "loss": 0.144, |
| "step": 244 |
| }, |
| { |
| "epoch": 9.8, |
| "learning_rate": 0.00019713966284282678, |
| "loss": 0.1692, |
| "step": 245 |
| }, |
| { |
| "epoch": 9.84, |
| "learning_rate": 0.0001971002663122945, |
| "loss": 0.1455, |
| "step": 246 |
| }, |
| { |
| "epoch": 9.88, |
| "learning_rate": 0.00019706060431334187, |
| "loss": 0.1591, |
| "step": 247 |
| }, |
| { |
| "epoch": 9.92, |
| "learning_rate": 0.00019702067695440332, |
| "loss": 0.2316, |
| "step": 248 |
| }, |
| { |
| "epoch": 9.96, |
| "learning_rate": 0.0001969804843446387, |
| "loss": 0.1383, |
| "step": 249 |
| }, |
| { |
| "epoch": 10.0, |
| "learning_rate": 0.00019694002659393305, |
| "loss": 0.1582, |
| "step": 250 |
| }, |
| { |
| "epoch": 10.04, |
| "learning_rate": 0.00019689930381289634, |
| "loss": 0.1053, |
| "step": 251 |
| }, |
| { |
| "epoch": 10.08, |
| "learning_rate": 0.0001968583161128631, |
| "loss": 0.1267, |
| "step": 252 |
| }, |
| { |
| "epoch": 10.12, |
| "learning_rate": 0.00019681706360589216, |
| "loss": 0.1441, |
| "step": 253 |
| }, |
| { |
| "epoch": 10.16, |
| "learning_rate": 0.00019677554640476624, |
| "loss": 0.0944, |
| "step": 254 |
| }, |
| { |
| "epoch": 10.2, |
| "learning_rate": 0.00019673376462299184, |
| "loss": 0.1168, |
| "step": 255 |
| }, |
| { |
| "epoch": 10.24, |
| "learning_rate": 0.00019669171837479873, |
| "loss": 0.0918, |
| "step": 256 |
| }, |
| { |
| "epoch": 10.28, |
| "learning_rate": 0.00019664940777513974, |
| "loss": 0.1484, |
| "step": 257 |
| }, |
| { |
| "epoch": 10.32, |
| "learning_rate": 0.00019660683293969041, |
| "loss": 0.092, |
| "step": 258 |
| }, |
| { |
| "epoch": 10.36, |
| "learning_rate": 0.0001965639939848488, |
| "loss": 0.179, |
| "step": 259 |
| }, |
| { |
| "epoch": 10.4, |
| "learning_rate": 0.00019652089102773488, |
| "loss": 0.1303, |
| "step": 260 |
| }, |
| { |
| "epoch": 10.44, |
| "learning_rate": 0.00019647752418619054, |
| "loss": 0.1263, |
| "step": 261 |
| }, |
| { |
| "epoch": 10.48, |
| "learning_rate": 0.00019643389357877907, |
| "loss": 0.1292, |
| "step": 262 |
| }, |
| { |
| "epoch": 10.52, |
| "learning_rate": 0.00019638999932478486, |
| "loss": 0.158, |
| "step": 263 |
| }, |
| { |
| "epoch": 10.56, |
| "learning_rate": 0.00019634584154421317, |
| "loss": 0.1215, |
| "step": 264 |
| }, |
| { |
| "epoch": 10.6, |
| "learning_rate": 0.00019630142035778964, |
| "loss": 0.1484, |
| "step": 265 |
| }, |
| { |
| "epoch": 10.64, |
| "learning_rate": 0.00019625673588696008, |
| "loss": 0.1418, |
| "step": 266 |
| }, |
| { |
| "epoch": 10.68, |
| "learning_rate": 0.0001962117882538902, |
| "loss": 0.1497, |
| "step": 267 |
| }, |
| { |
| "epoch": 10.72, |
| "learning_rate": 0.00019616657758146503, |
| "loss": 0.1388, |
| "step": 268 |
| }, |
| { |
| "epoch": 10.76, |
| "learning_rate": 0.00019612110399328892, |
| "loss": 0.1335, |
| "step": 269 |
| }, |
| { |
| "epoch": 10.8, |
| "learning_rate": 0.00019607536761368484, |
| "loss": 0.1513, |
| "step": 270 |
| }, |
| { |
| "epoch": 10.84, |
| "learning_rate": 0.0001960293685676943, |
| "loss": 0.1302, |
| "step": 271 |
| }, |
| { |
| "epoch": 10.88, |
| "learning_rate": 0.00019598310698107702, |
| "loss": 0.154, |
| "step": 272 |
| }, |
| { |
| "epoch": 10.92, |
| "learning_rate": 0.00019593658298031034, |
| "loss": 0.1187, |
| "step": 273 |
| }, |
| { |
| "epoch": 10.96, |
| "learning_rate": 0.0001958897966925891, |
| "loss": 0.1413, |
| "step": 274 |
| }, |
| { |
| "epoch": 11.0, |
| "learning_rate": 0.0001958427482458253, |
| "loss": 0.1538, |
| "step": 275 |
| }, |
| { |
| "epoch": 11.04, |
| "learning_rate": 0.0001957954377686475, |
| "loss": 0.0878, |
| "step": 276 |
| }, |
| { |
| "epoch": 11.08, |
| "learning_rate": 0.00019574786539040077, |
| "loss": 0.0819, |
| "step": 277 |
| }, |
| { |
| "epoch": 11.12, |
| "learning_rate": 0.00019570003124114619, |
| "loss": 0.1119, |
| "step": 278 |
| }, |
| { |
| "epoch": 11.16, |
| "learning_rate": 0.00019565193545166052, |
| "loss": 0.0974, |
| "step": 279 |
| }, |
| { |
| "epoch": 11.2, |
| "learning_rate": 0.00019560357815343577, |
| "loss": 0.15, |
| "step": 280 |
| }, |
| { |
| "epoch": 11.24, |
| "learning_rate": 0.00019555495947867895, |
| "loss": 0.1226, |
| "step": 281 |
| }, |
| { |
| "epoch": 11.28, |
| "learning_rate": 0.0001955060795603117, |
| "loss": 0.0949, |
| "step": 282 |
| }, |
| { |
| "epoch": 11.32, |
| "learning_rate": 0.00019545693853196983, |
| "loss": 0.0981, |
| "step": 283 |
| }, |
| { |
| "epoch": 11.36, |
| "learning_rate": 0.000195407536528003, |
| "loss": 0.0858, |
| "step": 284 |
| }, |
| { |
| "epoch": 11.4, |
| "learning_rate": 0.00019535787368347442, |
| "loss": 0.1373, |
| "step": 285 |
| }, |
| { |
| "epoch": 11.44, |
| "learning_rate": 0.00019530795013416046, |
| "loss": 0.1411, |
| "step": 286 |
| }, |
| { |
| "epoch": 11.48, |
| "learning_rate": 0.00019525776601655014, |
| "loss": 0.1739, |
| "step": 287 |
| }, |
| { |
| "epoch": 11.52, |
| "learning_rate": 0.00019520732146784491, |
| "loss": 0.1095, |
| "step": 288 |
| }, |
| { |
| "epoch": 11.56, |
| "learning_rate": 0.0001951566166259583, |
| "loss": 0.139, |
| "step": 289 |
| }, |
| { |
| "epoch": 11.6, |
| "learning_rate": 0.00019510565162951537, |
| "loss": 0.1112, |
| "step": 290 |
| }, |
| { |
| "epoch": 11.64, |
| "learning_rate": 0.0001950544266178525, |
| "loss": 0.1344, |
| "step": 291 |
| }, |
| { |
| "epoch": 11.68, |
| "learning_rate": 0.00019500294173101687, |
| "loss": 0.1556, |
| "step": 292 |
| }, |
| { |
| "epoch": 11.72, |
| "learning_rate": 0.00019495119710976626, |
| "loss": 0.1368, |
| "step": 293 |
| }, |
| { |
| "epoch": 11.76, |
| "learning_rate": 0.00019489919289556845, |
| "loss": 0.128, |
| "step": 294 |
| }, |
| { |
| "epoch": 11.8, |
| "learning_rate": 0.00019484692923060095, |
| "loss": 0.1357, |
| "step": 295 |
| }, |
| { |
| "epoch": 11.84, |
| "learning_rate": 0.0001947944062577507, |
| "loss": 0.1554, |
| "step": 296 |
| }, |
| { |
| "epoch": 11.88, |
| "learning_rate": 0.0001947416241206134, |
| "loss": 0.1415, |
| "step": 297 |
| }, |
| { |
| "epoch": 11.92, |
| "learning_rate": 0.0001946885829634935, |
| "loss": 0.1495, |
| "step": 298 |
| }, |
| { |
| "epoch": 11.96, |
| "learning_rate": 0.00019463528293140345, |
| "loss": 0.1326, |
| "step": 299 |
| }, |
| { |
| "epoch": 12.0, |
| "learning_rate": 0.00019458172417006347, |
| "loss": 0.1727, |
| "step": 300 |
| }, |
| { |
| "epoch": 12.04, |
| "learning_rate": 0.00019452790682590124, |
| "loss": 0.0949, |
| "step": 301 |
| }, |
| { |
| "epoch": 12.08, |
| "learning_rate": 0.00019447383104605125, |
| "loss": 0.1019, |
| "step": 302 |
| }, |
| { |
| "epoch": 12.12, |
| "learning_rate": 0.0001944194969783547, |
| "loss": 0.1067, |
| "step": 303 |
| }, |
| { |
| "epoch": 12.16, |
| "learning_rate": 0.00019436490477135878, |
| "loss": 0.1237, |
| "step": 304 |
| }, |
| { |
| "epoch": 12.2, |
| "learning_rate": 0.00019431005457431653, |
| "loss": 0.1031, |
| "step": 305 |
| }, |
| { |
| "epoch": 12.24, |
| "learning_rate": 0.0001942549465371863, |
| "loss": 0.116, |
| "step": 306 |
| }, |
| { |
| "epoch": 12.28, |
| "learning_rate": 0.00019419958081063138, |
| "loss": 0.1247, |
| "step": 307 |
| }, |
| { |
| "epoch": 12.32, |
| "learning_rate": 0.00019414395754601947, |
| "loss": 0.1112, |
| "step": 308 |
| }, |
| { |
| "epoch": 12.36, |
| "learning_rate": 0.00019408807689542257, |
| "loss": 0.1203, |
| "step": 309 |
| }, |
| { |
| "epoch": 12.4, |
| "learning_rate": 0.00019403193901161613, |
| "loss": 0.1154, |
| "step": 310 |
| }, |
| { |
| "epoch": 12.44, |
| "learning_rate": 0.00019397554404807906, |
| "loss": 0.0904, |
| "step": 311 |
| }, |
| { |
| "epoch": 12.48, |
| "learning_rate": 0.00019391889215899299, |
| "loss": 0.1156, |
| "step": 312 |
| }, |
| { |
| "epoch": 12.52, |
| "learning_rate": 0.00019386198349924207, |
| "loss": 0.125, |
| "step": 313 |
| }, |
| { |
| "epoch": 12.56, |
| "learning_rate": 0.00019380481822441235, |
| "loss": 0.1377, |
| "step": 314 |
| }, |
| { |
| "epoch": 12.6, |
| "learning_rate": 0.00019374739649079153, |
| "loss": 0.1014, |
| "step": 315 |
| }, |
| { |
| "epoch": 12.64, |
| "learning_rate": 0.00019368971845536845, |
| "loss": 0.1541, |
| "step": 316 |
| }, |
| { |
| "epoch": 12.68, |
| "learning_rate": 0.00019363178427583266, |
| "loss": 0.136, |
| "step": 317 |
| }, |
| { |
| "epoch": 12.72, |
| "learning_rate": 0.000193573594110574, |
| "loss": 0.1139, |
| "step": 318 |
| }, |
| { |
| "epoch": 12.76, |
| "learning_rate": 0.00019351514811868207, |
| "loss": 0.1742, |
| "step": 319 |
| }, |
| { |
| "epoch": 12.8, |
| "learning_rate": 0.0001934564464599461, |
| "loss": 0.1249, |
| "step": 320 |
| }, |
| { |
| "epoch": 12.84, |
| "learning_rate": 0.0001933974892948541, |
| "loss": 0.1236, |
| "step": 321 |
| }, |
| { |
| "epoch": 12.88, |
| "learning_rate": 0.0001933382767845928, |
| "loss": 0.1016, |
| "step": 322 |
| }, |
| { |
| "epoch": 12.92, |
| "learning_rate": 0.00019327880909104683, |
| "loss": 0.1095, |
| "step": 323 |
| }, |
| { |
| "epoch": 12.96, |
| "learning_rate": 0.00019321908637679865, |
| "loss": 0.1419, |
| "step": 324 |
| }, |
| { |
| "epoch": 13.0, |
| "learning_rate": 0.0001931591088051279, |
| "loss": 0.159, |
| "step": 325 |
| }, |
| { |
| "epoch": 13.04, |
| "learning_rate": 0.00019309887654001096, |
| "loss": 0.0903, |
| "step": 326 |
| }, |
| { |
| "epoch": 13.08, |
| "learning_rate": 0.0001930383897461205, |
| "loss": 0.0973, |
| "step": 327 |
| }, |
| { |
| "epoch": 13.12, |
| "learning_rate": 0.00019297764858882514, |
| "loss": 0.0843, |
| "step": 328 |
| }, |
| { |
| "epoch": 13.16, |
| "learning_rate": 0.0001929166532341889, |
| "loss": 0.0855, |
| "step": 329 |
| }, |
| { |
| "epoch": 13.2, |
| "learning_rate": 0.00019285540384897073, |
| "loss": 0.1237, |
| "step": 330 |
| }, |
| { |
| "epoch": 13.24, |
| "learning_rate": 0.00019279390060062407, |
| "loss": 0.1005, |
| "step": 331 |
| }, |
| { |
| "epoch": 13.28, |
| "learning_rate": 0.00019273214365729655, |
| "loss": 0.0814, |
| "step": 332 |
| }, |
| { |
| "epoch": 13.32, |
| "learning_rate": 0.0001926701331878292, |
| "loss": 0.1157, |
| "step": 333 |
| }, |
| { |
| "epoch": 13.36, |
| "learning_rate": 0.00019260786936175635, |
| "loss": 0.1184, |
| "step": 334 |
| }, |
| { |
| "epoch": 13.4, |
| "learning_rate": 0.00019254535234930486, |
| "loss": 0.1214, |
| "step": 335 |
| }, |
| { |
| "epoch": 13.44, |
| "learning_rate": 0.00019248258232139388, |
| "loss": 0.128, |
| "step": 336 |
| }, |
| { |
| "epoch": 13.48, |
| "learning_rate": 0.0001924195594496343, |
| "loss": 0.0972, |
| "step": 337 |
| }, |
| { |
| "epoch": 13.52, |
| "learning_rate": 0.00019235628390632822, |
| "loss": 0.0915, |
| "step": 338 |
| }, |
| { |
| "epoch": 13.56, |
| "learning_rate": 0.0001922927558644685, |
| "loss": 0.1229, |
| "step": 339 |
| }, |
| { |
| "epoch": 13.6, |
| "learning_rate": 0.00019222897549773848, |
| "loss": 0.109, |
| "step": 340 |
| }, |
| { |
| "epoch": 13.64, |
| "learning_rate": 0.0001921649429805112, |
| "loss": 0.1288, |
| "step": 341 |
| }, |
| { |
| "epoch": 13.68, |
| "learning_rate": 0.00019210065848784913, |
| "loss": 0.1264, |
| "step": 342 |
| }, |
| { |
| "epoch": 13.72, |
| "learning_rate": 0.00019203612219550358, |
| "loss": 0.1318, |
| "step": 343 |
| }, |
| { |
| "epoch": 13.76, |
| "learning_rate": 0.00019197133427991436, |
| "loss": 0.1657, |
| "step": 344 |
| }, |
| { |
| "epoch": 13.8, |
| "learning_rate": 0.00019190629491820912, |
| "loss": 0.1035, |
| "step": 345 |
| }, |
| { |
| "epoch": 13.84, |
| "learning_rate": 0.000191841004288203, |
| "loss": 0.1456, |
| "step": 346 |
| }, |
| { |
| "epoch": 13.88, |
| "learning_rate": 0.00019177546256839812, |
| "loss": 0.1385, |
| "step": 347 |
| }, |
| { |
| "epoch": 13.92, |
| "learning_rate": 0.000191709669937983, |
| "loss": 0.142, |
| "step": 348 |
| }, |
| { |
| "epoch": 13.96, |
| "learning_rate": 0.00019164362657683222, |
| "loss": 0.1301, |
| "step": 349 |
| }, |
| { |
| "epoch": 14.0, |
| "learning_rate": 0.00019157733266550575, |
| "loss": 0.1144, |
| "step": 350 |
| }, |
| { |
| "epoch": 14.04, |
| "learning_rate": 0.00019151078838524867, |
| "loss": 0.0813, |
| "step": 351 |
| }, |
| { |
| "epoch": 14.08, |
| "learning_rate": 0.00019144399391799043, |
| "loss": 0.0886, |
| "step": 352 |
| }, |
| { |
| "epoch": 14.12, |
| "learning_rate": 0.00019137694944634464, |
| "loss": 0.1104, |
| "step": 353 |
| }, |
| { |
| "epoch": 14.16, |
| "learning_rate": 0.0001913096551536083, |
| "loss": 0.1081, |
| "step": 354 |
| }, |
| { |
| "epoch": 14.2, |
| "learning_rate": 0.00019124211122376137, |
| "loss": 0.1039, |
| "step": 355 |
| }, |
| { |
| "epoch": 14.24, |
| "learning_rate": 0.00019117431784146645, |
| "loss": 0.1219, |
| "step": 356 |
| }, |
| { |
| "epoch": 14.28, |
| "learning_rate": 0.00019110627519206805, |
| "loss": 0.0936, |
| "step": 357 |
| }, |
| { |
| "epoch": 14.32, |
| "learning_rate": 0.00019103798346159213, |
| "loss": 0.1271, |
| "step": 358 |
| }, |
| { |
| "epoch": 14.36, |
| "learning_rate": 0.00019096944283674571, |
| "loss": 0.1121, |
| "step": 359 |
| }, |
| { |
| "epoch": 14.4, |
| "learning_rate": 0.00019090065350491626, |
| "loss": 0.0968, |
| "step": 360 |
| }, |
| { |
| "epoch": 14.44, |
| "learning_rate": 0.00019083161565417115, |
| "loss": 0.1063, |
| "step": 361 |
| }, |
| { |
| "epoch": 14.48, |
| "learning_rate": 0.00019076232947325722, |
| "loss": 0.1012, |
| "step": 362 |
| }, |
| { |
| "epoch": 14.52, |
| "learning_rate": 0.00019069279515160025, |
| "loss": 0.0927, |
| "step": 363 |
| }, |
| { |
| "epoch": 14.56, |
| "learning_rate": 0.00019062301287930446, |
| "loss": 0.1253, |
| "step": 364 |
| }, |
| { |
| "epoch": 14.6, |
| "learning_rate": 0.00019055298284715192, |
| "loss": 0.1191, |
| "step": 365 |
| }, |
| { |
| "epoch": 14.64, |
| "learning_rate": 0.00019048270524660196, |
| "loss": 0.1187, |
| "step": 366 |
| }, |
| { |
| "epoch": 14.68, |
| "learning_rate": 0.00019041218026979095, |
| "loss": 0.163, |
| "step": 367 |
| }, |
| { |
| "epoch": 14.72, |
| "learning_rate": 0.0001903414081095315, |
| "loss": 0.1191, |
| "step": 368 |
| }, |
| { |
| "epoch": 14.76, |
| "learning_rate": 0.0001902703889593119, |
| "loss": 0.1118, |
| "step": 369 |
| }, |
| { |
| "epoch": 14.8, |
| "learning_rate": 0.00019019912301329592, |
| "loss": 0.0916, |
| "step": 370 |
| }, |
| { |
| "epoch": 14.84, |
| "learning_rate": 0.0001901276104663218, |
| "loss": 0.1047, |
| "step": 371 |
| }, |
| { |
| "epoch": 14.88, |
| "learning_rate": 0.00019005585151390223, |
| "loss": 0.1085, |
| "step": 372 |
| }, |
| { |
| "epoch": 14.92, |
| "learning_rate": 0.00018998384635222334, |
| "loss": 0.1198, |
| "step": 373 |
| }, |
| { |
| "epoch": 14.96, |
| "learning_rate": 0.0001899115951781446, |
| "loss": 0.1693, |
| "step": 374 |
| }, |
| { |
| "epoch": 15.0, |
| "learning_rate": 0.0001898390981891979, |
| "loss": 0.1007, |
| "step": 375 |
| }, |
| { |
| "epoch": 15.04, |
| "learning_rate": 0.00018976635558358722, |
| "loss": 0.0896, |
| "step": 376 |
| }, |
| { |
| "epoch": 15.08, |
| "learning_rate": 0.00018969336756018815, |
| "loss": 0.0948, |
| "step": 377 |
| }, |
| { |
| "epoch": 15.12, |
| "learning_rate": 0.00018962013431854702, |
| "loss": 0.0894, |
| "step": 378 |
| }, |
| { |
| "epoch": 15.16, |
| "learning_rate": 0.00018954665605888088, |
| "loss": 0.0979, |
| "step": 379 |
| }, |
| { |
| "epoch": 15.2, |
| "learning_rate": 0.00018947293298207635, |
| "loss": 0.1138, |
| "step": 380 |
| }, |
| { |
| "epoch": 15.24, |
| "learning_rate": 0.0001893989652896896, |
| "loss": 0.0881, |
| "step": 381 |
| }, |
| { |
| "epoch": 15.28, |
| "learning_rate": 0.0001893247531839454, |
| "loss": 0.1252, |
| "step": 382 |
| }, |
| { |
| "epoch": 15.32, |
| "learning_rate": 0.0001892502968677369, |
| "loss": 0.0904, |
| "step": 383 |
| }, |
| { |
| "epoch": 15.36, |
| "learning_rate": 0.00018917559654462474, |
| "loss": 0.1047, |
| "step": 384 |
| }, |
| { |
| "epoch": 15.4, |
| "learning_rate": 0.0001891006524188368, |
| "loss": 0.1169, |
| "step": 385 |
| }, |
| { |
| "epoch": 15.44, |
| "learning_rate": 0.00018902546469526743, |
| "loss": 0.0865, |
| "step": 386 |
| }, |
| { |
| "epoch": 15.48, |
| "learning_rate": 0.00018895003357947705, |
| "loss": 0.1157, |
| "step": 387 |
| }, |
| { |
| "epoch": 15.52, |
| "learning_rate": 0.00018887435927769137, |
| "loss": 0.0881, |
| "step": 388 |
| }, |
| { |
| "epoch": 15.56, |
| "learning_rate": 0.0001887984419968011, |
| "loss": 0.1007, |
| "step": 389 |
| }, |
| { |
| "epoch": 15.6, |
| "learning_rate": 0.0001887222819443612, |
| "loss": 0.1201, |
| "step": 390 |
| }, |
| { |
| "epoch": 15.64, |
| "learning_rate": 0.00018864587932859028, |
| "loss": 0.0859, |
| "step": 391 |
| }, |
| { |
| "epoch": 15.68, |
| "learning_rate": 0.00018856923435837022, |
| "loss": 0.1152, |
| "step": 392 |
| }, |
| { |
| "epoch": 15.72, |
| "learning_rate": 0.00018849234724324543, |
| "loss": 0.1245, |
| "step": 393 |
| }, |
| { |
| "epoch": 15.76, |
| "learning_rate": 0.00018841521819342236, |
| "loss": 0.1018, |
| "step": 394 |
| }, |
| { |
| "epoch": 15.8, |
| "learning_rate": 0.0001883378474197689, |
| "loss": 0.116, |
| "step": 395 |
| }, |
| { |
| "epoch": 15.84, |
| "learning_rate": 0.0001882602351338137, |
| "loss": 0.1032, |
| "step": 396 |
| }, |
| { |
| "epoch": 15.88, |
| "learning_rate": 0.00018818238154774588, |
| "loss": 0.1424, |
| "step": 397 |
| }, |
| { |
| "epoch": 15.92, |
| "learning_rate": 0.00018810428687441414, |
| "loss": 0.1298, |
| "step": 398 |
| }, |
| { |
| "epoch": 15.96, |
| "learning_rate": 0.00018802595132732635, |
| "loss": 0.1012, |
| "step": 399 |
| }, |
| { |
| "epoch": 16.0, |
| "learning_rate": 0.0001879473751206489, |
| "loss": 0.1427, |
| "step": 400 |
| }, |
| { |
| "epoch": 16.04, |
| "learning_rate": 0.00018786855846920615, |
| "loss": 0.0966, |
| "step": 401 |
| }, |
| { |
| "epoch": 16.08, |
| "learning_rate": 0.00018778950158847976, |
| "loss": 0.0967, |
| "step": 402 |
| }, |
| { |
| "epoch": 16.12, |
| "learning_rate": 0.0001877102046946083, |
| "loss": 0.1336, |
| "step": 403 |
| }, |
| { |
| "epoch": 16.16, |
| "learning_rate": 0.00018763066800438636, |
| "loss": 0.0841, |
| "step": 404 |
| }, |
| { |
| "epoch": 16.2, |
| "learning_rate": 0.0001875508917352643, |
| "loss": 0.0968, |
| "step": 405 |
| }, |
| { |
| "epoch": 16.24, |
| "learning_rate": 0.00018747087610534736, |
| "loss": 0.0954, |
| "step": 406 |
| }, |
| { |
| "epoch": 16.28, |
| "learning_rate": 0.00018739062133339517, |
| "loss": 0.0946, |
| "step": 407 |
| }, |
| { |
| "epoch": 16.32, |
| "learning_rate": 0.00018731012763882133, |
| "loss": 0.1232, |
| "step": 408 |
| }, |
| { |
| "epoch": 16.36, |
| "learning_rate": 0.0001872293952416924, |
| "loss": 0.0716, |
| "step": 409 |
| }, |
| { |
| "epoch": 16.4, |
| "learning_rate": 0.00018714842436272773, |
| "loss": 0.157, |
| "step": 410 |
| }, |
| { |
| "epoch": 16.44, |
| "learning_rate": 0.00018706721522329862, |
| "loss": 0.0939, |
| "step": 411 |
| }, |
| { |
| "epoch": 16.48, |
| "learning_rate": 0.00018698576804542777, |
| "loss": 0.1066, |
| "step": 412 |
| }, |
| { |
| "epoch": 16.52, |
| "learning_rate": 0.00018690408305178856, |
| "loss": 0.1035, |
| "step": 413 |
| }, |
| { |
| "epoch": 16.56, |
| "learning_rate": 0.00018682216046570475, |
| "loss": 0.1175, |
| "step": 414 |
| }, |
| { |
| "epoch": 16.6, |
| "learning_rate": 0.00018674000051114952, |
| "loss": 0.1123, |
| "step": 415 |
| }, |
| { |
| "epoch": 16.64, |
| "learning_rate": 0.00018665760341274505, |
| "loss": 0.1069, |
| "step": 416 |
| }, |
| { |
| "epoch": 16.68, |
| "learning_rate": 0.00018657496939576186, |
| "loss": 0.1021, |
| "step": 417 |
| }, |
| { |
| "epoch": 16.72, |
| "learning_rate": 0.0001864920986861182, |
| "loss": 0.1102, |
| "step": 418 |
| }, |
| { |
| "epoch": 16.76, |
| "learning_rate": 0.00018640899151037944, |
| "loss": 0.1202, |
| "step": 419 |
| }, |
| { |
| "epoch": 16.8, |
| "learning_rate": 0.00018632564809575742, |
| "loss": 0.112, |
| "step": 420 |
| }, |
| { |
| "epoch": 16.84, |
| "learning_rate": 0.0001862420686701098, |
| "loss": 0.1061, |
| "step": 421 |
| }, |
| { |
| "epoch": 16.88, |
| "learning_rate": 0.0001861582534619396, |
| "loss": 0.1123, |
| "step": 422 |
| }, |
| { |
| "epoch": 16.92, |
| "learning_rate": 0.0001860742027003944, |
| "loss": 0.1051, |
| "step": 423 |
| }, |
| { |
| "epoch": 16.96, |
| "learning_rate": 0.00018598991661526572, |
| "loss": 0.1303, |
| "step": 424 |
| }, |
| { |
| "epoch": 17.0, |
| "learning_rate": 0.00018590539543698854, |
| "loss": 0.0821, |
| "step": 425 |
| }, |
| { |
| "epoch": 17.04, |
| "learning_rate": 0.0001858206393966405, |
| "loss": 0.1033, |
| "step": 426 |
| }, |
| { |
| "epoch": 17.08, |
| "learning_rate": 0.00018573564872594145, |
| "loss": 0.1269, |
| "step": 427 |
| }, |
| { |
| "epoch": 17.12, |
| "learning_rate": 0.00018565042365725258, |
| "loss": 0.0999, |
| "step": 428 |
| }, |
| { |
| "epoch": 17.16, |
| "learning_rate": 0.00018556496442357595, |
| "loss": 0.1045, |
| "step": 429 |
| }, |
| { |
| "epoch": 17.2, |
| "learning_rate": 0.0001854792712585539, |
| "loss": 0.0921, |
| "step": 430 |
| }, |
| { |
| "epoch": 17.24, |
| "learning_rate": 0.00018539334439646824, |
| "loss": 0.0733, |
| "step": 431 |
| }, |
| { |
| "epoch": 17.28, |
| "learning_rate": 0.00018530718407223974, |
| "loss": 0.0993, |
| "step": 432 |
| }, |
| { |
| "epoch": 17.32, |
| "learning_rate": 0.00018522079052142747, |
| "loss": 0.1052, |
| "step": 433 |
| }, |
| { |
| "epoch": 17.36, |
| "learning_rate": 0.00018513416398022802, |
| "loss": 0.108, |
| "step": 434 |
| }, |
| { |
| "epoch": 17.4, |
| "learning_rate": 0.0001850473046854751, |
| "loss": 0.1008, |
| "step": 435 |
| }, |
| { |
| "epoch": 17.44, |
| "learning_rate": 0.0001849602128746387, |
| "loss": 0.0913, |
| "step": 436 |
| }, |
| { |
| "epoch": 17.48, |
| "learning_rate": 0.00018487288878582447, |
| "loss": 0.1079, |
| "step": 437 |
| }, |
| { |
| "epoch": 17.52, |
| "learning_rate": 0.00018478533265777318, |
| "loss": 0.0877, |
| "step": 438 |
| }, |
| { |
| "epoch": 17.56, |
| "learning_rate": 0.00018469754472985993, |
| "loss": 0.082, |
| "step": 439 |
| }, |
| { |
| "epoch": 17.6, |
| "learning_rate": 0.00018460952524209355, |
| "loss": 0.1091, |
| "step": 440 |
| }, |
| { |
| "epoch": 17.64, |
| "learning_rate": 0.00018452127443511598, |
| "loss": 0.085, |
| "step": 441 |
| }, |
| { |
| "epoch": 17.68, |
| "learning_rate": 0.00018443279255020152, |
| "loss": 0.0975, |
| "step": 442 |
| }, |
| { |
| "epoch": 17.72, |
| "learning_rate": 0.0001843440798292563, |
| "loss": 0.0968, |
| "step": 443 |
| }, |
| { |
| "epoch": 17.76, |
| "learning_rate": 0.00018425513651481747, |
| "loss": 0.0837, |
| "step": 444 |
| }, |
| { |
| "epoch": 17.8, |
| "learning_rate": 0.00018416596285005272, |
| "loss": 0.1137, |
| "step": 445 |
| }, |
| { |
| "epoch": 17.84, |
| "learning_rate": 0.0001840765590787594, |
| "loss": 0.0942, |
| "step": 446 |
| }, |
| { |
| "epoch": 17.88, |
| "learning_rate": 0.00018398692544536397, |
| "loss": 0.12, |
| "step": 447 |
| }, |
| { |
| "epoch": 17.92, |
| "learning_rate": 0.00018389706219492147, |
| "loss": 0.0947, |
| "step": 448 |
| }, |
| { |
| "epoch": 17.96, |
| "learning_rate": 0.00018380696957311449, |
| "loss": 0.1327, |
| "step": 449 |
| }, |
| { |
| "epoch": 18.0, |
| "learning_rate": 0.00018371664782625287, |
| "loss": 0.0958, |
| "step": 450 |
| }, |
| { |
| "epoch": 18.04, |
| "learning_rate": 0.0001836260972012728, |
| "loss": 0.0929, |
| "step": 451 |
| }, |
| { |
| "epoch": 18.08, |
| "learning_rate": 0.00018353531794573625, |
| "loss": 0.0726, |
| "step": 452 |
| }, |
| { |
| "epoch": 18.12, |
| "learning_rate": 0.0001834443103078302, |
| "loss": 0.0928, |
| "step": 453 |
| }, |
| { |
| "epoch": 18.16, |
| "learning_rate": 0.0001833530745363661, |
| "loss": 0.0728, |
| "step": 454 |
| }, |
| { |
| "epoch": 18.2, |
| "learning_rate": 0.00018326161088077903, |
| "loss": 0.077, |
| "step": 455 |
| }, |
| { |
| "epoch": 18.24, |
| "learning_rate": 0.00018316991959112716, |
| "loss": 0.0846, |
| "step": 456 |
| }, |
| { |
| "epoch": 18.28, |
| "learning_rate": 0.00018307800091809097, |
| "loss": 0.0924, |
| "step": 457 |
| }, |
| { |
| "epoch": 18.32, |
| "learning_rate": 0.0001829858551129726, |
| "loss": 0.1072, |
| "step": 458 |
| }, |
| { |
| "epoch": 18.36, |
| "learning_rate": 0.00018289348242769515, |
| "loss": 0.0853, |
| "step": 459 |
| }, |
| { |
| "epoch": 18.4, |
| "learning_rate": 0.00018280088311480201, |
| "loss": 0.1256, |
| "step": 460 |
| }, |
| { |
| "epoch": 18.44, |
| "learning_rate": 0.00018270805742745617, |
| "loss": 0.1144, |
| "step": 461 |
| }, |
| { |
| "epoch": 18.48, |
| "learning_rate": 0.00018261500561943955, |
| "loss": 0.0946, |
| "step": 462 |
| }, |
| { |
| "epoch": 18.52, |
| "learning_rate": 0.00018252172794515223, |
| "loss": 0.0913, |
| "step": 463 |
| }, |
| { |
| "epoch": 18.56, |
| "learning_rate": 0.00018242822465961176, |
| "loss": 0.0848, |
| "step": 464 |
| }, |
| { |
| "epoch": 18.6, |
| "learning_rate": 0.00018233449601845258, |
| "loss": 0.1579, |
| "step": 465 |
| }, |
| { |
| "epoch": 18.64, |
| "learning_rate": 0.00018224054227792524, |
| "loss": 0.0876, |
| "step": 466 |
| }, |
| { |
| "epoch": 18.68, |
| "learning_rate": 0.00018214636369489563, |
| "loss": 0.1041, |
| "step": 467 |
| }, |
| { |
| "epoch": 18.72, |
| "learning_rate": 0.00018205196052684445, |
| "loss": 0.1219, |
| "step": 468 |
| }, |
| { |
| "epoch": 18.76, |
| "learning_rate": 0.00018195733303186633, |
| "loss": 0.0987, |
| "step": 469 |
| }, |
| { |
| "epoch": 18.8, |
| "learning_rate": 0.00018186248146866927, |
| "loss": 0.0936, |
| "step": 470 |
| }, |
| { |
| "epoch": 18.84, |
| "learning_rate": 0.0001817674060965737, |
| "loss": 0.0878, |
| "step": 471 |
| }, |
| { |
| "epoch": 18.88, |
| "learning_rate": 0.00018167210717551224, |
| "loss": 0.1141, |
| "step": 472 |
| }, |
| { |
| "epoch": 18.92, |
| "learning_rate": 0.00018157658496602833, |
| "loss": 0.1005, |
| "step": 473 |
| }, |
| { |
| "epoch": 18.96, |
| "learning_rate": 0.00018148083972927616, |
| "loss": 0.1682, |
| "step": 474 |
| }, |
| { |
| "epoch": 19.0, |
| "learning_rate": 0.0001813848717270195, |
| "loss": 0.1425, |
| "step": 475 |
| }, |
| { |
| "epoch": 19.04, |
| "learning_rate": 0.00018128868122163123, |
| "loss": 0.0882, |
| "step": 476 |
| }, |
| { |
| "epoch": 19.08, |
| "learning_rate": 0.00018119226847609245, |
| "loss": 0.0923, |
| "step": 477 |
| }, |
| { |
| "epoch": 19.12, |
| "learning_rate": 0.000181095633753992, |
| "loss": 0.1027, |
| "step": 478 |
| }, |
| { |
| "epoch": 19.16, |
| "learning_rate": 0.0001809987773195255, |
| "loss": 0.082, |
| "step": 479 |
| }, |
| { |
| "epoch": 19.2, |
| "learning_rate": 0.00018090169943749476, |
| "loss": 0.0767, |
| "step": 480 |
| }, |
| { |
| "epoch": 19.24, |
| "learning_rate": 0.00018080440037330695, |
| "loss": 0.0908, |
| "step": 481 |
| }, |
| { |
| "epoch": 19.28, |
| "learning_rate": 0.00018070688039297403, |
| "loss": 0.1036, |
| "step": 482 |
| }, |
| { |
| "epoch": 19.32, |
| "learning_rate": 0.00018060913976311192, |
| "loss": 0.1075, |
| "step": 483 |
| }, |
| { |
| "epoch": 19.36, |
| "learning_rate": 0.00018051117875093976, |
| "loss": 0.0882, |
| "step": 484 |
| }, |
| { |
| "epoch": 19.4, |
| "learning_rate": 0.00018041299762427916, |
| "loss": 0.0947, |
| "step": 485 |
| }, |
| { |
| "epoch": 19.44, |
| "learning_rate": 0.00018031459665155363, |
| "loss": 0.0851, |
| "step": 486 |
| }, |
| { |
| "epoch": 19.48, |
| "learning_rate": 0.00018021597610178768, |
| "loss": 0.0987, |
| "step": 487 |
| }, |
| { |
| "epoch": 19.52, |
| "learning_rate": 0.00018011713624460608, |
| "loss": 0.0873, |
| "step": 488 |
| }, |
| { |
| "epoch": 19.56, |
| "learning_rate": 0.0001800180773502333, |
| "loss": 0.1079, |
| "step": 489 |
| }, |
| { |
| "epoch": 19.6, |
| "learning_rate": 0.0001799187996894925, |
| "loss": 0.0978, |
| "step": 490 |
| }, |
| { |
| "epoch": 19.64, |
| "learning_rate": 0.00017981930353380503, |
| "loss": 0.0977, |
| "step": 491 |
| }, |
| { |
| "epoch": 19.68, |
| "learning_rate": 0.0001797195891551896, |
| "loss": 0.081, |
| "step": 492 |
| }, |
| { |
| "epoch": 19.72, |
| "learning_rate": 0.0001796196568262615, |
| "loss": 0.1058, |
| "step": 493 |
| }, |
| { |
| "epoch": 19.76, |
| "learning_rate": 0.00017951950682023191, |
| "loss": 0.0825, |
| "step": 494 |
| }, |
| { |
| "epoch": 19.8, |
| "learning_rate": 0.0001794191394109071, |
| "loss": 0.0844, |
| "step": 495 |
| }, |
| { |
| "epoch": 19.84, |
| "learning_rate": 0.00017931855487268782, |
| "loss": 0.0979, |
| "step": 496 |
| }, |
| { |
| "epoch": 19.88, |
| "learning_rate": 0.0001792177534805682, |
| "loss": 0.1559, |
| "step": 497 |
| }, |
| { |
| "epoch": 19.92, |
| "learning_rate": 0.00017911673551013551, |
| "loss": 0.106, |
| "step": 498 |
| }, |
| { |
| "epoch": 19.96, |
| "learning_rate": 0.00017901550123756906, |
| "loss": 0.1059, |
| "step": 499 |
| }, |
| { |
| "epoch": 20.0, |
| "learning_rate": 0.00017891405093963938, |
| "loss": 0.1164, |
| "step": 500 |
| }, |
| { |
| "epoch": 20.04, |
| "learning_rate": 0.00017881238489370776, |
| "loss": 0.1031, |
| "step": 501 |
| }, |
| { |
| "epoch": 20.08, |
| "learning_rate": 0.00017871050337772525, |
| "loss": 0.0924, |
| "step": 502 |
| }, |
| { |
| "epoch": 20.12, |
| "learning_rate": 0.00017860840667023212, |
| "loss": 0.0655, |
| "step": 503 |
| }, |
| { |
| "epoch": 20.16, |
| "learning_rate": 0.0001785060950503568, |
| "loss": 0.0988, |
| "step": 504 |
| }, |
| { |
| "epoch": 20.2, |
| "learning_rate": 0.0001784035687978153, |
| "loss": 0.0785, |
| "step": 505 |
| }, |
| { |
| "epoch": 20.24, |
| "learning_rate": 0.0001783008281929106, |
| "loss": 0.0898, |
| "step": 506 |
| }, |
| { |
| "epoch": 20.28, |
| "learning_rate": 0.0001781978735165315, |
| "loss": 0.0878, |
| "step": 507 |
| }, |
| { |
| "epoch": 20.32, |
| "learning_rate": 0.0001780947050501522, |
| "loss": 0.0987, |
| "step": 508 |
| }, |
| { |
| "epoch": 20.36, |
| "learning_rate": 0.00017799132307583132, |
| "loss": 0.1007, |
| "step": 509 |
| }, |
| { |
| "epoch": 20.4, |
| "learning_rate": 0.00017788772787621126, |
| "loss": 0.1026, |
| "step": 510 |
| }, |
| { |
| "epoch": 20.44, |
| "learning_rate": 0.0001777839197345173, |
| "loss": 0.0834, |
| "step": 511 |
| }, |
| { |
| "epoch": 20.48, |
| "learning_rate": 0.00017767989893455698, |
| "loss": 0.0949, |
| "step": 512 |
| }, |
| { |
| "epoch": 20.52, |
| "learning_rate": 0.00017757566576071914, |
| "loss": 0.0666, |
| "step": 513 |
| }, |
| { |
| "epoch": 20.56, |
| "learning_rate": 0.00017747122049797335, |
| "loss": 0.1034, |
| "step": 514 |
| }, |
| { |
| "epoch": 20.6, |
| "learning_rate": 0.00017736656343186896, |
| "loss": 0.0821, |
| "step": 515 |
| }, |
| { |
| "epoch": 20.64, |
| "learning_rate": 0.00017726169484853438, |
| "loss": 0.1007, |
| "step": 516 |
| }, |
| { |
| "epoch": 20.68, |
| "learning_rate": 0.0001771566150346763, |
| "loss": 0.0674, |
| "step": 517 |
| }, |
| { |
| "epoch": 20.72, |
| "learning_rate": 0.00017705132427757895, |
| "loss": 0.0916, |
| "step": 518 |
| }, |
| { |
| "epoch": 20.76, |
| "learning_rate": 0.0001769458228651032, |
| "loss": 0.107, |
| "step": 519 |
| }, |
| { |
| "epoch": 20.8, |
| "learning_rate": 0.00017684011108568592, |
| "loss": 0.0835, |
| "step": 520 |
| }, |
| { |
| "epoch": 20.84, |
| "learning_rate": 0.00017673418922833903, |
| "loss": 0.0998, |
| "step": 521 |
| }, |
| { |
| "epoch": 20.88, |
| "learning_rate": 0.00017662805758264893, |
| "loss": 0.0897, |
| "step": 522 |
| }, |
| { |
| "epoch": 20.92, |
| "learning_rate": 0.0001765217164387754, |
| "loss": 0.1198, |
| "step": 523 |
| }, |
| { |
| "epoch": 20.96, |
| "learning_rate": 0.00017641516608745114, |
| "loss": 0.1038, |
| "step": 524 |
| }, |
| { |
| "epoch": 21.0, |
| "learning_rate": 0.00017630840681998066, |
| "loss": 0.0816, |
| "step": 525 |
| }, |
| { |
| "epoch": 21.04, |
| "learning_rate": 0.00017620143892823977, |
| "loss": 0.0752, |
| "step": 526 |
| }, |
| { |
| "epoch": 21.08, |
| "learning_rate": 0.00017609426270467462, |
| "loss": 0.0962, |
| "step": 527 |
| }, |
| { |
| "epoch": 21.12, |
| "learning_rate": 0.00017598687844230088, |
| "loss": 0.0784, |
| "step": 528 |
| }, |
| { |
| "epoch": 21.16, |
| "learning_rate": 0.000175879286434703, |
| "loss": 0.0911, |
| "step": 529 |
| }, |
| { |
| "epoch": 21.2, |
| "learning_rate": 0.0001757714869760335, |
| "loss": 0.1017, |
| "step": 530 |
| }, |
| { |
| "epoch": 21.24, |
| "learning_rate": 0.00017566348036101187, |
| "loss": 0.083, |
| "step": 531 |
| }, |
| { |
| "epoch": 21.28, |
| "learning_rate": 0.0001755552668849242, |
| "loss": 0.0994, |
| "step": 532 |
| }, |
| { |
| "epoch": 21.32, |
| "learning_rate": 0.0001754468468436219, |
| "loss": 0.0877, |
| "step": 533 |
| }, |
| { |
| "epoch": 21.36, |
| "learning_rate": 0.00017533822053352128, |
| "loss": 0.0896, |
| "step": 534 |
| }, |
| { |
| "epoch": 21.4, |
| "learning_rate": 0.0001752293882516025, |
| "loss": 0.0756, |
| "step": 535 |
| }, |
| { |
| "epoch": 21.44, |
| "learning_rate": 0.00017512035029540885, |
| "loss": 0.087, |
| "step": 536 |
| }, |
| { |
| "epoch": 21.48, |
| "learning_rate": 0.00017501110696304596, |
| "loss": 0.0933, |
| "step": 537 |
| }, |
| { |
| "epoch": 21.52, |
| "learning_rate": 0.00017490165855318094, |
| "loss": 0.0833, |
| "step": 538 |
| }, |
| { |
| "epoch": 21.56, |
| "learning_rate": 0.00017479200536504156, |
| "loss": 0.105, |
| "step": 539 |
| }, |
| { |
| "epoch": 21.6, |
| "learning_rate": 0.0001746821476984154, |
| "loss": 0.1003, |
| "step": 540 |
| }, |
| { |
| "epoch": 21.64, |
| "learning_rate": 0.00017457208585364918, |
| "loss": 0.0702, |
| "step": 541 |
| }, |
| { |
| "epoch": 21.68, |
| "learning_rate": 0.00017446182013164778, |
| "loss": 0.0867, |
| "step": 542 |
| }, |
| { |
| "epoch": 21.72, |
| "learning_rate": 0.00017435135083387345, |
| "loss": 0.09, |
| "step": 543 |
| }, |
| { |
| "epoch": 21.76, |
| "learning_rate": 0.000174240678262345, |
| "loss": 0.0939, |
| "step": 544 |
| }, |
| { |
| "epoch": 21.8, |
| "learning_rate": 0.0001741298027196371, |
| "loss": 0.0809, |
| "step": 545 |
| }, |
| { |
| "epoch": 21.84, |
| "learning_rate": 0.00017401872450887917, |
| "loss": 0.1442, |
| "step": 546 |
| }, |
| { |
| "epoch": 21.88, |
| "learning_rate": 0.00017390744393375486, |
| "loss": 0.0936, |
| "step": 547 |
| }, |
| { |
| "epoch": 21.92, |
| "learning_rate": 0.00017379596129850098, |
| "loss": 0.1098, |
| "step": 548 |
| }, |
| { |
| "epoch": 21.96, |
| "learning_rate": 0.00017368427690790677, |
| "loss": 0.0734, |
| "step": 549 |
| }, |
| { |
| "epoch": 22.0, |
| "learning_rate": 0.00017357239106731317, |
| "loss": 0.1212, |
| "step": 550 |
| }, |
| { |
| "epoch": 22.04, |
| "learning_rate": 0.00017346030408261172, |
| "loss": 0.0887, |
| "step": 551 |
| }, |
| { |
| "epoch": 22.08, |
| "learning_rate": 0.000173348016260244, |
| "loss": 0.0897, |
| "step": 552 |
| }, |
| { |
| "epoch": 22.12, |
| "learning_rate": 0.00017323552790720058, |
| "loss": 0.0852, |
| "step": 553 |
| }, |
| { |
| "epoch": 22.16, |
| "learning_rate": 0.00017312283933102038, |
| "loss": 0.0797, |
| "step": 554 |
| }, |
| { |
| "epoch": 22.2, |
| "learning_rate": 0.00017300995083978965, |
| "loss": 0.063, |
| "step": 555 |
| }, |
| { |
| "epoch": 22.24, |
| "learning_rate": 0.00017289686274214118, |
| "loss": 0.1112, |
| "step": 556 |
| }, |
| { |
| "epoch": 22.28, |
| "learning_rate": 0.0001727835753472535, |
| "loss": 0.1024, |
| "step": 557 |
| }, |
| { |
| "epoch": 22.32, |
| "learning_rate": 0.0001726700889648501, |
| "loss": 0.0801, |
| "step": 558 |
| }, |
| { |
| "epoch": 22.36, |
| "learning_rate": 0.00017255640390519836, |
| "loss": 0.0848, |
| "step": 559 |
| }, |
| { |
| "epoch": 22.4, |
| "learning_rate": 0.00017244252047910892, |
| "loss": 0.0852, |
| "step": 560 |
| }, |
| { |
| "epoch": 22.44, |
| "learning_rate": 0.00017232843899793468, |
| "loss": 0.0898, |
| "step": 561 |
| }, |
| { |
| "epoch": 22.48, |
| "learning_rate": 0.00017221415977357007, |
| "loss": 0.0889, |
| "step": 562 |
| }, |
| { |
| "epoch": 22.52, |
| "learning_rate": 0.00017209968311845012, |
| "loss": 0.0957, |
| "step": 563 |
| }, |
| { |
| "epoch": 22.56, |
| "learning_rate": 0.00017198500934554966, |
| "loss": 0.0905, |
| "step": 564 |
| }, |
| { |
| "epoch": 22.6, |
| "learning_rate": 0.0001718701387683824, |
| "loss": 0.0792, |
| "step": 565 |
| }, |
| { |
| "epoch": 22.64, |
| "learning_rate": 0.0001717550717010001, |
| "loss": 0.0849, |
| "step": 566 |
| }, |
| { |
| "epoch": 22.68, |
| "learning_rate": 0.0001716398084579917, |
| "loss": 0.0781, |
| "step": 567 |
| }, |
| { |
| "epoch": 22.72, |
| "learning_rate": 0.00017152434935448256, |
| "loss": 0.0893, |
| "step": 568 |
| }, |
| { |
| "epoch": 22.76, |
| "learning_rate": 0.00017140869470613342, |
| "loss": 0.0788, |
| "step": 569 |
| }, |
| { |
| "epoch": 22.8, |
| "learning_rate": 0.00017129284482913972, |
| "loss": 0.086, |
| "step": 570 |
| }, |
| { |
| "epoch": 22.84, |
| "learning_rate": 0.00017117680004023056, |
| "loss": 0.0927, |
| "step": 571 |
| }, |
| { |
| "epoch": 22.88, |
| "learning_rate": 0.00017106056065666793, |
| "loss": 0.0768, |
| "step": 572 |
| }, |
| { |
| "epoch": 22.92, |
| "learning_rate": 0.00017094412699624595, |
| "loss": 0.0841, |
| "step": 573 |
| }, |
| { |
| "epoch": 22.96, |
| "learning_rate": 0.00017082749937728973, |
| "loss": 0.0918, |
| "step": 574 |
| }, |
| { |
| "epoch": 23.0, |
| "learning_rate": 0.00017071067811865476, |
| "loss": 0.1138, |
| "step": 575 |
| }, |
| { |
| "epoch": 23.04, |
| "learning_rate": 0.0001705936635397259, |
| "loss": 0.0807, |
| "step": 576 |
| }, |
| { |
| "epoch": 23.08, |
| "learning_rate": 0.00017047645596041653, |
| "loss": 0.0852, |
| "step": 577 |
| }, |
| { |
| "epoch": 23.12, |
| "learning_rate": 0.0001703590557011677, |
| "loss": 0.0681, |
| "step": 578 |
| }, |
| { |
| "epoch": 23.16, |
| "learning_rate": 0.00017024146308294724, |
| "loss": 0.0768, |
| "step": 579 |
| }, |
| { |
| "epoch": 23.2, |
| "learning_rate": 0.00017012367842724887, |
| "loss": 0.0678, |
| "step": 580 |
| }, |
| { |
| "epoch": 23.24, |
| "learning_rate": 0.00017000570205609136, |
| "loss": 0.0803, |
| "step": 581 |
| }, |
| { |
| "epoch": 23.28, |
| "learning_rate": 0.00016988753429201755, |
| "loss": 0.0804, |
| "step": 582 |
| }, |
| { |
| "epoch": 23.32, |
| "learning_rate": 0.00016976917545809367, |
| "loss": 0.1088, |
| "step": 583 |
| }, |
| { |
| "epoch": 23.36, |
| "learning_rate": 0.00016965062587790823, |
| "loss": 0.0774, |
| "step": 584 |
| }, |
| { |
| "epoch": 23.4, |
| "learning_rate": 0.00016953188587557122, |
| "loss": 0.0731, |
| "step": 585 |
| }, |
| { |
| "epoch": 23.44, |
| "learning_rate": 0.0001694129557757133, |
| "loss": 0.0774, |
| "step": 586 |
| }, |
| { |
| "epoch": 23.48, |
| "learning_rate": 0.0001692938359034848, |
| "loss": 0.0781, |
| "step": 587 |
| }, |
| { |
| "epoch": 23.52, |
| "learning_rate": 0.00016917452658455495, |
| "loss": 0.0889, |
| "step": 588 |
| }, |
| { |
| "epoch": 23.56, |
| "learning_rate": 0.00016905502814511082, |
| "loss": 0.1092, |
| "step": 589 |
| }, |
| { |
| "epoch": 23.6, |
| "learning_rate": 0.0001689353409118566, |
| "loss": 0.0844, |
| "step": 590 |
| }, |
| { |
| "epoch": 23.64, |
| "learning_rate": 0.0001688154652120126, |
| "loss": 0.0858, |
| "step": 591 |
| }, |
| { |
| "epoch": 23.68, |
| "learning_rate": 0.00016869540137331445, |
| "loss": 0.0999, |
| "step": 592 |
| }, |
| { |
| "epoch": 23.72, |
| "learning_rate": 0.00016857514972401207, |
| "loss": 0.0633, |
| "step": 593 |
| }, |
| { |
| "epoch": 23.76, |
| "learning_rate": 0.00016845471059286887, |
| "loss": 0.07, |
| "step": 594 |
| }, |
| { |
| "epoch": 23.8, |
| "learning_rate": 0.00016833408430916085, |
| "loss": 0.0797, |
| "step": 595 |
| }, |
| { |
| "epoch": 23.84, |
| "learning_rate": 0.00016821327120267567, |
| "loss": 0.0911, |
| "step": 596 |
| }, |
| { |
| "epoch": 23.88, |
| "learning_rate": 0.0001680922716037117, |
| "loss": 0.0694, |
| "step": 597 |
| }, |
| { |
| "epoch": 23.92, |
| "learning_rate": 0.00016797108584307732, |
| "loss": 0.0826, |
| "step": 598 |
| }, |
| { |
| "epoch": 23.96, |
| "learning_rate": 0.00016784971425208965, |
| "loss": 0.1116, |
| "step": 599 |
| }, |
| { |
| "epoch": 24.0, |
| "learning_rate": 0.00016772815716257412, |
| "loss": 0.0987, |
| "step": 600 |
| }, |
| { |
| "epoch": 24.04, |
| "learning_rate": 0.00016760641490686307, |
| "loss": 0.0714, |
| "step": 601 |
| }, |
| { |
| "epoch": 24.08, |
| "learning_rate": 0.0001674844878177952, |
| "loss": 0.0854, |
| "step": 602 |
| }, |
| { |
| "epoch": 24.12, |
| "learning_rate": 0.00016736237622871452, |
| "loss": 0.0796, |
| "step": 603 |
| }, |
| { |
| "epoch": 24.16, |
| "learning_rate": 0.00016724008047346947, |
| "loss": 0.0568, |
| "step": 604 |
| }, |
| { |
| "epoch": 24.2, |
| "learning_rate": 0.00016711760088641196, |
| "loss": 0.0677, |
| "step": 605 |
| }, |
| { |
| "epoch": 24.24, |
| "learning_rate": 0.0001669949378023965, |
| "loss": 0.0701, |
| "step": 606 |
| }, |
| { |
| "epoch": 24.28, |
| "learning_rate": 0.00016687209155677929, |
| "loss": 0.0658, |
| "step": 607 |
| }, |
| { |
| "epoch": 24.32, |
| "learning_rate": 0.00016674906248541726, |
| "loss": 0.0794, |
| "step": 608 |
| }, |
| { |
| "epoch": 24.36, |
| "learning_rate": 0.00016662585092466723, |
| "loss": 0.0742, |
| "step": 609 |
| }, |
| { |
| "epoch": 24.4, |
| "learning_rate": 0.0001665024572113848, |
| "loss": 0.0657, |
| "step": 610 |
| }, |
| { |
| "epoch": 24.44, |
| "learning_rate": 0.00016637888168292384, |
| "loss": 0.0613, |
| "step": 611 |
| }, |
| { |
| "epoch": 24.48, |
| "learning_rate": 0.000166255124677135, |
| "loss": 0.0949, |
| "step": 612 |
| }, |
| { |
| "epoch": 24.52, |
| "learning_rate": 0.00016613118653236518, |
| "loss": 0.0802, |
| "step": 613 |
| }, |
| { |
| "epoch": 24.56, |
| "learning_rate": 0.00016600706758745668, |
| "loss": 0.0933, |
| "step": 614 |
| }, |
| { |
| "epoch": 24.6, |
| "learning_rate": 0.0001658827681817458, |
| "loss": 0.0743, |
| "step": 615 |
| }, |
| { |
| "epoch": 24.64, |
| "learning_rate": 0.00016575828865506245, |
| "loss": 0.0831, |
| "step": 616 |
| }, |
| { |
| "epoch": 24.68, |
| "learning_rate": 0.00016563362934772892, |
| "loss": 0.0722, |
| "step": 617 |
| }, |
| { |
| "epoch": 24.72, |
| "learning_rate": 0.00016550879060055895, |
| "loss": 0.0846, |
| "step": 618 |
| }, |
| { |
| "epoch": 24.76, |
| "learning_rate": 0.00016538377275485691, |
| "loss": 0.0798, |
| "step": 619 |
| }, |
| { |
| "epoch": 24.8, |
| "learning_rate": 0.00016525857615241687, |
| "loss": 0.0842, |
| "step": 620 |
| }, |
| { |
| "epoch": 24.84, |
| "learning_rate": 0.00016513320113552152, |
| "loss": 0.1014, |
| "step": 621 |
| }, |
| { |
| "epoch": 24.88, |
| "learning_rate": 0.0001650076480469413, |
| "loss": 0.0727, |
| "step": 622 |
| }, |
| { |
| "epoch": 24.92, |
| "learning_rate": 0.0001648819172299337, |
| "loss": 0.0921, |
| "step": 623 |
| }, |
| { |
| "epoch": 24.96, |
| "learning_rate": 0.0001647560090282419, |
| "loss": 0.0793, |
| "step": 624 |
| }, |
| { |
| "epoch": 25.0, |
| "learning_rate": 0.00016462992378609407, |
| "loss": 0.0817, |
| "step": 625 |
| }, |
| { |
| "epoch": 25.04, |
| "learning_rate": 0.00016450366184820255, |
| "loss": 0.0821, |
| "step": 626 |
| }, |
| { |
| "epoch": 25.08, |
| "learning_rate": 0.00016437722355976258, |
| "loss": 0.0611, |
| "step": 627 |
| }, |
| { |
| "epoch": 25.12, |
| "learning_rate": 0.00016425060926645167, |
| "loss": 0.071, |
| "step": 628 |
| }, |
| { |
| "epoch": 25.16, |
| "learning_rate": 0.00016412381931442838, |
| "loss": 0.0653, |
| "step": 629 |
| }, |
| { |
| "epoch": 25.2, |
| "learning_rate": 0.00016399685405033167, |
| "loss": 0.0855, |
| "step": 630 |
| }, |
| { |
| "epoch": 25.24, |
| "learning_rate": 0.0001638697138212797, |
| "loss": 0.0632, |
| "step": 631 |
| }, |
| { |
| "epoch": 25.28, |
| "learning_rate": 0.000163742398974869, |
| "loss": 0.0834, |
| "step": 632 |
| }, |
| { |
| "epoch": 25.32, |
| "learning_rate": 0.0001636149098591735, |
| "loss": 0.0906, |
| "step": 633 |
| }, |
| { |
| "epoch": 25.36, |
| "learning_rate": 0.00016348724682274353, |
| "loss": 0.0795, |
| "step": 634 |
| }, |
| { |
| "epoch": 25.4, |
| "learning_rate": 0.00016335941021460506, |
| "loss": 0.0817, |
| "step": 635 |
| }, |
| { |
| "epoch": 25.44, |
| "learning_rate": 0.00016323140038425842, |
| "loss": 0.0803, |
| "step": 636 |
| }, |
| { |
| "epoch": 25.48, |
| "learning_rate": 0.00016310321768167762, |
| "loss": 0.0877, |
| "step": 637 |
| }, |
| { |
| "epoch": 25.52, |
| "learning_rate": 0.00016297486245730927, |
| "loss": 0.084, |
| "step": 638 |
| }, |
| { |
| "epoch": 25.56, |
| "learning_rate": 0.0001628463350620716, |
| "loss": 0.0623, |
| "step": 639 |
| }, |
| { |
| "epoch": 25.6, |
| "learning_rate": 0.0001627176358473537, |
| "loss": 0.0841, |
| "step": 640 |
| }, |
| { |
| "epoch": 25.64, |
| "learning_rate": 0.00016258876516501424, |
| "loss": 0.0599, |
| "step": 641 |
| }, |
| { |
| "epoch": 25.68, |
| "learning_rate": 0.0001624597233673808, |
| "loss": 0.0659, |
| "step": 642 |
| }, |
| { |
| "epoch": 25.72, |
| "learning_rate": 0.00016233051080724868, |
| "loss": 0.0874, |
| "step": 643 |
| }, |
| { |
| "epoch": 25.76, |
| "learning_rate": 0.0001622011278378801, |
| "loss": 0.0744, |
| "step": 644 |
| }, |
| { |
| "epoch": 25.8, |
| "learning_rate": 0.00016207157481300312, |
| "loss": 0.0591, |
| "step": 645 |
| }, |
| { |
| "epoch": 25.84, |
| "learning_rate": 0.00016194185208681083, |
| "loss": 0.0843, |
| "step": 646 |
| }, |
| { |
| "epoch": 25.88, |
| "learning_rate": 0.00016181196001396019, |
| "loss": 0.0876, |
| "step": 647 |
| }, |
| { |
| "epoch": 25.92, |
| "learning_rate": 0.0001616818989495711, |
| "loss": 0.0957, |
| "step": 648 |
| }, |
| { |
| "epoch": 25.96, |
| "learning_rate": 0.00016155166924922566, |
| "loss": 0.0678, |
| "step": 649 |
| }, |
| { |
| "epoch": 26.0, |
| "learning_rate": 0.0001614212712689668, |
| "loss": 0.0732, |
| "step": 650 |
| }, |
| { |
| "epoch": 26.04, |
| "learning_rate": 0.00016129070536529766, |
| "loss": 0.073, |
| "step": 651 |
| }, |
| { |
| "epoch": 26.08, |
| "learning_rate": 0.00016115997189518043, |
| "loss": 0.0763, |
| "step": 652 |
| }, |
| { |
| "epoch": 26.12, |
| "learning_rate": 0.00016102907121603543, |
| "loss": 0.0583, |
| "step": 653 |
| }, |
| { |
| "epoch": 26.16, |
| "learning_rate": 0.00016089800368574014, |
| "loss": 0.0716, |
| "step": 654 |
| }, |
| { |
| "epoch": 26.2, |
| "learning_rate": 0.00016076676966262813, |
| "loss": 0.0742, |
| "step": 655 |
| }, |
| { |
| "epoch": 26.24, |
| "learning_rate": 0.00016063536950548826, |
| "loss": 0.0719, |
| "step": 656 |
| }, |
| { |
| "epoch": 26.28, |
| "learning_rate": 0.0001605038035735635, |
| "loss": 0.056, |
| "step": 657 |
| }, |
| { |
| "epoch": 26.32, |
| "learning_rate": 0.0001603720722265501, |
| "loss": 0.0693, |
| "step": 658 |
| }, |
| { |
| "epoch": 26.36, |
| "learning_rate": 0.00016024017582459652, |
| "loss": 0.0548, |
| "step": 659 |
| }, |
| { |
| "epoch": 26.4, |
| "learning_rate": 0.00016010811472830252, |
| "loss": 0.0713, |
| "step": 660 |
| }, |
| { |
| "epoch": 26.44, |
| "learning_rate": 0.00015997588929871808, |
| "loss": 0.0737, |
| "step": 661 |
| }, |
| { |
| "epoch": 26.48, |
| "learning_rate": 0.00015984349989734247, |
| "loss": 0.084, |
| "step": 662 |
| }, |
| { |
| "epoch": 26.52, |
| "learning_rate": 0.00015971094688612322, |
| "loss": 0.0709, |
| "step": 663 |
| }, |
| { |
| "epoch": 26.56, |
| "learning_rate": 0.0001595782306274553, |
| "loss": 0.0708, |
| "step": 664 |
| }, |
| { |
| "epoch": 26.6, |
| "learning_rate": 0.00015944535148417982, |
| "loss": 0.0904, |
| "step": 665 |
| }, |
| { |
| "epoch": 26.64, |
| "learning_rate": 0.00015931230981958326, |
| "loss": 0.0649, |
| "step": 666 |
| }, |
| { |
| "epoch": 26.68, |
| "learning_rate": 0.0001591791059973965, |
| "loss": 0.0762, |
| "step": 667 |
| }, |
| { |
| "epoch": 26.72, |
| "learning_rate": 0.0001590457403817937, |
| "loss": 0.0623, |
| "step": 668 |
| }, |
| { |
| "epoch": 26.76, |
| "learning_rate": 0.00015891221333739136, |
| "loss": 0.0839, |
| "step": 669 |
| }, |
| { |
| "epoch": 26.8, |
| "learning_rate": 0.00015877852522924732, |
| "loss": 0.1045, |
| "step": 670 |
| }, |
| { |
| "epoch": 26.84, |
| "learning_rate": 0.00015864467642285975, |
| "loss": 0.0725, |
| "step": 671 |
| }, |
| { |
| "epoch": 26.88, |
| "learning_rate": 0.00015851066728416618, |
| "loss": 0.0821, |
| "step": 672 |
| }, |
| { |
| "epoch": 26.92, |
| "learning_rate": 0.00015837649817954246, |
| "loss": 0.0844, |
| "step": 673 |
| }, |
| { |
| "epoch": 26.96, |
| "learning_rate": 0.00015824216947580183, |
| "loss": 0.0881, |
| "step": 674 |
| }, |
| { |
| "epoch": 27.0, |
| "learning_rate": 0.00015810768154019385, |
| "loss": 0.0805, |
| "step": 675 |
| }, |
| { |
| "epoch": 27.04, |
| "learning_rate": 0.00015797303474040332, |
| "loss": 0.0649, |
| "step": 676 |
| }, |
| { |
| "epoch": 27.08, |
| "learning_rate": 0.00015783822944454956, |
| "loss": 0.0684, |
| "step": 677 |
| }, |
| { |
| "epoch": 27.12, |
| "learning_rate": 0.000157703266021185, |
| "loss": 0.0685, |
| "step": 678 |
| }, |
| { |
| "epoch": 27.16, |
| "learning_rate": 0.00015756814483929458, |
| "loss": 0.0662, |
| "step": 679 |
| }, |
| { |
| "epoch": 27.2, |
| "learning_rate": 0.00015743286626829437, |
| "loss": 0.0803, |
| "step": 680 |
| }, |
| { |
| "epoch": 27.24, |
| "learning_rate": 0.00015729743067803087, |
| "loss": 0.0663, |
| "step": 681 |
| }, |
| { |
| "epoch": 27.28, |
| "learning_rate": 0.00015716183843877976, |
| "loss": 0.0639, |
| "step": 682 |
| }, |
| { |
| "epoch": 27.32, |
| "learning_rate": 0.0001570260899212451, |
| "loss": 0.0622, |
| "step": 683 |
| }, |
| { |
| "epoch": 27.36, |
| "learning_rate": 0.00015689018549655813, |
| "loss": 0.0571, |
| "step": 684 |
| }, |
| { |
| "epoch": 27.4, |
| "learning_rate": 0.00015675412553627639, |
| "loss": 0.0997, |
| "step": 685 |
| }, |
| { |
| "epoch": 27.44, |
| "learning_rate": 0.00015661791041238256, |
| "loss": 0.1829, |
| "step": 686 |
| }, |
| { |
| "epoch": 27.48, |
| "learning_rate": 0.00015648154049728358, |
| "loss": 0.0797, |
| "step": 687 |
| }, |
| { |
| "epoch": 27.52, |
| "learning_rate": 0.00015634501616380967, |
| "loss": 0.0585, |
| "step": 688 |
| }, |
| { |
| "epoch": 27.56, |
| "learning_rate": 0.00015620833778521307, |
| "loss": 0.0698, |
| "step": 689 |
| }, |
| { |
| "epoch": 27.6, |
| "learning_rate": 0.0001560715057351673, |
| "loss": 0.0848, |
| "step": 690 |
| }, |
| { |
| "epoch": 27.64, |
| "learning_rate": 0.0001559345203877659, |
| "loss": 0.0744, |
| "step": 691 |
| }, |
| { |
| "epoch": 27.68, |
| "learning_rate": 0.00015579738211752165, |
| "loss": 0.08, |
| "step": 692 |
| }, |
| { |
| "epoch": 27.72, |
| "learning_rate": 0.00015566009129936527, |
| "loss": 0.0505, |
| "step": 693 |
| }, |
| { |
| "epoch": 27.76, |
| "learning_rate": 0.00015552264830864468, |
| "loss": 0.0769, |
| "step": 694 |
| }, |
| { |
| "epoch": 27.8, |
| "learning_rate": 0.00015538505352112375, |
| "loss": 0.065, |
| "step": 695 |
| }, |
| { |
| "epoch": 27.84, |
| "learning_rate": 0.00015524730731298134, |
| "loss": 0.0665, |
| "step": 696 |
| }, |
| { |
| "epoch": 27.88, |
| "learning_rate": 0.00015510941006081032, |
| "loss": 0.0526, |
| "step": 697 |
| }, |
| { |
| "epoch": 27.92, |
| "learning_rate": 0.00015497136214161664, |
| "loss": 0.0767, |
| "step": 698 |
| }, |
| { |
| "epoch": 27.96, |
| "learning_rate": 0.00015483316393281784, |
| "loss": 0.0649, |
| "step": 699 |
| }, |
| { |
| "epoch": 28.0, |
| "learning_rate": 0.00015469481581224272, |
| "loss": 0.0603, |
| "step": 700 |
| }, |
| { |
| "epoch": 28.04, |
| "learning_rate": 0.00015455631815812962, |
| "loss": 0.0672, |
| "step": 701 |
| }, |
| { |
| "epoch": 28.08, |
| "learning_rate": 0.00015441767134912596, |
| "loss": 0.0531, |
| "step": 702 |
| }, |
| { |
| "epoch": 28.12, |
| "learning_rate": 0.00015427887576428674, |
| "loss": 0.094, |
| "step": 703 |
| }, |
| { |
| "epoch": 28.16, |
| "learning_rate": 0.0001541399317830738, |
| "loss": 0.0527, |
| "step": 704 |
| }, |
| { |
| "epoch": 28.2, |
| "learning_rate": 0.00015400083978535473, |
| "loss": 0.0621, |
| "step": 705 |
| }, |
| { |
| "epoch": 28.24, |
| "learning_rate": 0.00015386160015140168, |
| "loss": 0.0811, |
| "step": 706 |
| }, |
| { |
| "epoch": 28.28, |
| "learning_rate": 0.0001537222132618905, |
| "loss": 0.0611, |
| "step": 707 |
| }, |
| { |
| "epoch": 28.32, |
| "learning_rate": 0.00015358267949789966, |
| "loss": 0.0724, |
| "step": 708 |
| }, |
| { |
| "epoch": 28.36, |
| "learning_rate": 0.00015344299924090915, |
| "loss": 0.0755, |
| "step": 709 |
| }, |
| { |
| "epoch": 28.4, |
| "learning_rate": 0.0001533031728727994, |
| "loss": 0.0565, |
| "step": 710 |
| }, |
| { |
| "epoch": 28.44, |
| "learning_rate": 0.00015316320077585037, |
| "loss": 0.0822, |
| "step": 711 |
| }, |
| { |
| "epoch": 28.48, |
| "learning_rate": 0.0001530230833327405, |
| "loss": 0.0632, |
| "step": 712 |
| }, |
| { |
| "epoch": 28.52, |
| "learning_rate": 0.00015288282092654543, |
| "loss": 0.0572, |
| "step": 713 |
| }, |
| { |
| "epoch": 28.56, |
| "learning_rate": 0.00015274241394073733, |
| "loss": 0.0685, |
| "step": 714 |
| }, |
| { |
| "epoch": 28.6, |
| "learning_rate": 0.00015260186275918342, |
| "loss": 0.0691, |
| "step": 715 |
| }, |
| { |
| "epoch": 28.64, |
| "learning_rate": 0.00015246116776614538, |
| "loss": 0.0696, |
| "step": 716 |
| }, |
| { |
| "epoch": 28.68, |
| "learning_rate": 0.00015232032934627782, |
| "loss": 0.064, |
| "step": 717 |
| }, |
| { |
| "epoch": 28.72, |
| "learning_rate": 0.00015217934788462774, |
| "loss": 0.0697, |
| "step": 718 |
| }, |
| { |
| "epoch": 28.76, |
| "learning_rate": 0.000152038223766633, |
| "loss": 0.0656, |
| "step": 719 |
| }, |
| { |
| "epoch": 28.8, |
| "learning_rate": 0.00015189695737812152, |
| "loss": 0.0692, |
| "step": 720 |
| }, |
| { |
| "epoch": 28.84, |
| "learning_rate": 0.0001517555491053103, |
| "loss": 0.0736, |
| "step": 721 |
| }, |
| { |
| "epoch": 28.88, |
| "learning_rate": 0.00015161399933480402, |
| "loss": 0.0618, |
| "step": 722 |
| }, |
| { |
| "epoch": 28.92, |
| "learning_rate": 0.00015147230845359452, |
| "loss": 0.0703, |
| "step": 723 |
| }, |
| { |
| "epoch": 28.96, |
| "learning_rate": 0.00015133047684905916, |
| "loss": 0.0782, |
| "step": 724 |
| }, |
| { |
| "epoch": 29.0, |
| "learning_rate": 0.00015118850490896012, |
| "loss": 0.0684, |
| "step": 725 |
| }, |
| { |
| "epoch": 29.04, |
| "learning_rate": 0.00015104639302144327, |
| "loss": 0.052, |
| "step": 726 |
| }, |
| { |
| "epoch": 29.08, |
| "learning_rate": 0.00015090414157503714, |
| "loss": 0.0781, |
| "step": 727 |
| }, |
| { |
| "epoch": 29.12, |
| "learning_rate": 0.0001507617509586517, |
| "loss": 0.0551, |
| "step": 728 |
| }, |
| { |
| "epoch": 29.16, |
| "learning_rate": 0.00015061922156157746, |
| "loss": 0.0697, |
| "step": 729 |
| }, |
| { |
| "epoch": 29.2, |
| "learning_rate": 0.0001504765537734844, |
| "loss": 0.071, |
| "step": 730 |
| }, |
| { |
| "epoch": 29.24, |
| "learning_rate": 0.00015033374798442074, |
| "loss": 0.0615, |
| "step": 731 |
| }, |
| { |
| "epoch": 29.28, |
| "learning_rate": 0.00015019080458481202, |
| "loss": 0.077, |
| "step": 732 |
| }, |
| { |
| "epoch": 29.32, |
| "learning_rate": 0.0001500477239654602, |
| "loss": 0.0587, |
| "step": 733 |
| }, |
| { |
| "epoch": 29.36, |
| "learning_rate": 0.00014990450651754207, |
| "loss": 0.0628, |
| "step": 734 |
| }, |
| { |
| "epoch": 29.4, |
| "learning_rate": 0.00014976115263260875, |
| "loss": 0.0807, |
| "step": 735 |
| }, |
| { |
| "epoch": 29.44, |
| "learning_rate": 0.00014961766270258422, |
| "loss": 0.0592, |
| "step": 736 |
| }, |
| { |
| "epoch": 29.48, |
| "learning_rate": 0.00014947403711976454, |
| "loss": 0.0646, |
| "step": 737 |
| }, |
| { |
| "epoch": 29.52, |
| "learning_rate": 0.0001493302762768165, |
| "loss": 0.0635, |
| "step": 738 |
| }, |
| { |
| "epoch": 29.56, |
| "learning_rate": 0.0001491863805667768, |
| "loss": 0.0621, |
| "step": 739 |
| }, |
| { |
| "epoch": 29.6, |
| "learning_rate": 0.00014904235038305083, |
| "loss": 0.0688, |
| "step": 740 |
| }, |
| { |
| "epoch": 29.64, |
| "learning_rate": 0.00014889818611941156, |
| "loss": 0.0659, |
| "step": 741 |
| }, |
| { |
| "epoch": 29.68, |
| "learning_rate": 0.00014875388816999865, |
| "loss": 0.053, |
| "step": 742 |
| }, |
| { |
| "epoch": 29.72, |
| "learning_rate": 0.00014860945692931712, |
| "loss": 0.0573, |
| "step": 743 |
| }, |
| { |
| "epoch": 29.76, |
| "learning_rate": 0.00014846489279223652, |
| "loss": 0.0624, |
| "step": 744 |
| }, |
| { |
| "epoch": 29.8, |
| "learning_rate": 0.00014832019615398963, |
| "loss": 0.0597, |
| "step": 745 |
| }, |
| { |
| "epoch": 29.84, |
| "learning_rate": 0.00014817536741017152, |
| "loss": 0.0602, |
| "step": 746 |
| }, |
| { |
| "epoch": 29.88, |
| "learning_rate": 0.00014803040695673852, |
| "loss": 0.0596, |
| "step": 747 |
| }, |
| { |
| "epoch": 29.92, |
| "learning_rate": 0.00014788531519000696, |
| "loss": 0.0718, |
| "step": 748 |
| }, |
| { |
| "epoch": 29.96, |
| "learning_rate": 0.00014774009250665215, |
| "loss": 0.0793, |
| "step": 749 |
| }, |
| { |
| "epoch": 30.0, |
| "learning_rate": 0.00014759473930370736, |
| "loss": 0.0685, |
| "step": 750 |
| }, |
| { |
| "epoch": 30.04, |
| "learning_rate": 0.00014744925597856275, |
| "loss": 0.0664, |
| "step": 751 |
| }, |
| { |
| "epoch": 30.08, |
| "learning_rate": 0.0001473036429289641, |
| "loss": 0.06, |
| "step": 752 |
| }, |
| { |
| "epoch": 30.12, |
| "learning_rate": 0.00014715790055301196, |
| "loss": 0.0598, |
| "step": 753 |
| }, |
| { |
| "epoch": 30.16, |
| "learning_rate": 0.0001470120292491605, |
| "loss": 0.0495, |
| "step": 754 |
| }, |
| { |
| "epoch": 30.2, |
| "learning_rate": 0.00014686602941621615, |
| "loss": 0.0471, |
| "step": 755 |
| }, |
| { |
| "epoch": 30.24, |
| "learning_rate": 0.00014671990145333696, |
| "loss": 0.0563, |
| "step": 756 |
| }, |
| { |
| "epoch": 30.28, |
| "learning_rate": 0.00014657364576003122, |
| "loss": 0.0635, |
| "step": 757 |
| }, |
| { |
| "epoch": 30.32, |
| "learning_rate": 0.0001464272627361564, |
| "loss": 0.0782, |
| "step": 758 |
| }, |
| { |
| "epoch": 30.36, |
| "learning_rate": 0.000146280752781918, |
| "loss": 0.0701, |
| "step": 759 |
| }, |
| { |
| "epoch": 30.4, |
| "learning_rate": 0.0001461341162978688, |
| "loss": 0.0584, |
| "step": 760 |
| }, |
| { |
| "epoch": 30.44, |
| "learning_rate": 0.00014598735368490725, |
| "loss": 0.0621, |
| "step": 761 |
| }, |
| { |
| "epoch": 30.48, |
| "learning_rate": 0.0001458404653442767, |
| "loss": 0.0538, |
| "step": 762 |
| }, |
| { |
| "epoch": 30.52, |
| "learning_rate": 0.00014569345167756437, |
| "loss": 0.0514, |
| "step": 763 |
| }, |
| { |
| "epoch": 30.56, |
| "learning_rate": 0.00014554631308669994, |
| "loss": 0.0683, |
| "step": 764 |
| }, |
| { |
| "epoch": 30.6, |
| "learning_rate": 0.00014539904997395468, |
| "loss": 0.0705, |
| "step": 765 |
| }, |
| { |
| "epoch": 30.64, |
| "learning_rate": 0.00014525166274194037, |
| "loss": 0.0846, |
| "step": 766 |
| }, |
| { |
| "epoch": 30.68, |
| "learning_rate": 0.00014510415179360805, |
| "loss": 0.0682, |
| "step": 767 |
| }, |
| { |
| "epoch": 30.72, |
| "learning_rate": 0.00014495651753224705, |
| "loss": 0.0621, |
| "step": 768 |
| }, |
| { |
| "epoch": 30.76, |
| "learning_rate": 0.00014480876036148378, |
| "loss": 0.0617, |
| "step": 769 |
| }, |
| { |
| "epoch": 30.8, |
| "learning_rate": 0.00014466088068528068, |
| "loss": 0.0676, |
| "step": 770 |
| }, |
| { |
| "epoch": 30.84, |
| "learning_rate": 0.00014451287890793513, |
| "loss": 0.0697, |
| "step": 771 |
| }, |
| { |
| "epoch": 30.88, |
| "learning_rate": 0.00014436475543407843, |
| "loss": 0.0604, |
| "step": 772 |
| }, |
| { |
| "epoch": 30.92, |
| "learning_rate": 0.00014421651066867438, |
| "loss": 0.0695, |
| "step": 773 |
| }, |
| { |
| "epoch": 30.96, |
| "learning_rate": 0.00014406814501701857, |
| "loss": 0.071, |
| "step": 774 |
| }, |
| { |
| "epoch": 31.0, |
| "learning_rate": 0.00014391965888473703, |
| "loss": 0.0711, |
| "step": 775 |
| }, |
| { |
| "epoch": 31.04, |
| "learning_rate": 0.00014377105267778518, |
| "loss": 0.0563, |
| "step": 776 |
| }, |
| { |
| "epoch": 31.08, |
| "learning_rate": 0.00014362232680244672, |
| "loss": 0.0542, |
| "step": 777 |
| }, |
| { |
| "epoch": 31.12, |
| "learning_rate": 0.00014347348166533248, |
| "loss": 0.0772, |
| "step": 778 |
| }, |
| { |
| "epoch": 31.16, |
| "learning_rate": 0.00014332451767337938, |
| "loss": 0.0622, |
| "step": 779 |
| }, |
| { |
| "epoch": 31.2, |
| "learning_rate": 0.00014317543523384928, |
| "loss": 0.0686, |
| "step": 780 |
| }, |
| { |
| "epoch": 31.24, |
| "learning_rate": 0.00014302623475432794, |
| "loss": 0.0586, |
| "step": 781 |
| }, |
| { |
| "epoch": 31.28, |
| "learning_rate": 0.00014287691664272375, |
| "loss": 0.0585, |
| "step": 782 |
| }, |
| { |
| "epoch": 31.32, |
| "learning_rate": 0.00014272748130726668, |
| "loss": 0.077, |
| "step": 783 |
| }, |
| { |
| "epoch": 31.36, |
| "learning_rate": 0.00014257792915650728, |
| "loss": 0.061, |
| "step": 784 |
| }, |
| { |
| "epoch": 31.4, |
| "learning_rate": 0.00014242826059931537, |
| "loss": 0.061, |
| "step": 785 |
| }, |
| { |
| "epoch": 31.44, |
| "learning_rate": 0.00014227847604487913, |
| "loss": 0.058, |
| "step": 786 |
| }, |
| { |
| "epoch": 31.48, |
| "learning_rate": 0.0001421285759027038, |
| "loss": 0.0725, |
| "step": 787 |
| }, |
| { |
| "epoch": 31.52, |
| "learning_rate": 0.0001419785605826106, |
| "loss": 0.0622, |
| "step": 788 |
| }, |
| { |
| "epoch": 31.56, |
| "learning_rate": 0.00014182843049473568, |
| "loss": 0.063, |
| "step": 789 |
| }, |
| { |
| "epoch": 31.6, |
| "learning_rate": 0.00014167818604952906, |
| "loss": 0.0789, |
| "step": 790 |
| }, |
| { |
| "epoch": 31.64, |
| "learning_rate": 0.00014152782765775322, |
| "loss": 0.0601, |
| "step": 791 |
| }, |
| { |
| "epoch": 31.68, |
| "learning_rate": 0.00014137735573048233, |
| "loss": 0.0617, |
| "step": 792 |
| }, |
| { |
| "epoch": 31.72, |
| "learning_rate": 0.00014122677067910085, |
| "loss": 0.0749, |
| "step": 793 |
| }, |
| { |
| "epoch": 31.76, |
| "learning_rate": 0.00014107607291530256, |
| "loss": 0.0599, |
| "step": 794 |
| }, |
| { |
| "epoch": 31.8, |
| "learning_rate": 0.0001409252628510894, |
| "loss": 0.0576, |
| "step": 795 |
| }, |
| { |
| "epoch": 31.84, |
| "learning_rate": 0.00014077434089877037, |
| "loss": 0.0636, |
| "step": 796 |
| }, |
| { |
| "epoch": 31.88, |
| "learning_rate": 0.00014062330747096027, |
| "loss": 0.0597, |
| "step": 797 |
| }, |
| { |
| "epoch": 31.92, |
| "learning_rate": 0.00014047216298057873, |
| "loss": 0.0679, |
| "step": 798 |
| }, |
| { |
| "epoch": 31.96, |
| "learning_rate": 0.00014032090784084903, |
| "loss": 0.0776, |
| "step": 799 |
| }, |
| { |
| "epoch": 32.0, |
| "learning_rate": 0.00014016954246529696, |
| "loss": 0.0653, |
| "step": 800 |
| }, |
| { |
| "epoch": 32.04, |
| "learning_rate": 0.00014001806726774964, |
| "loss": 0.0572, |
| "step": 801 |
| }, |
| { |
| "epoch": 32.08, |
| "learning_rate": 0.00013986648266233452, |
| "loss": 0.0809, |
| "step": 802 |
| }, |
| { |
| "epoch": 32.12, |
| "learning_rate": 0.00013971478906347806, |
| "loss": 0.0575, |
| "step": 803 |
| }, |
| { |
| "epoch": 32.16, |
| "learning_rate": 0.00013956298688590484, |
| "loss": 0.0591, |
| "step": 804 |
| }, |
| { |
| "epoch": 32.2, |
| "learning_rate": 0.0001394110765446362, |
| "loss": 0.0633, |
| "step": 805 |
| }, |
| { |
| "epoch": 32.24, |
| "learning_rate": 0.00013925905845498914, |
| "loss": 0.0572, |
| "step": 806 |
| }, |
| { |
| "epoch": 32.28, |
| "learning_rate": 0.00013910693303257541, |
| "loss": 0.0545, |
| "step": 807 |
| }, |
| { |
| "epoch": 32.32, |
| "learning_rate": 0.00013895470069330004, |
| "loss": 0.0688, |
| "step": 808 |
| }, |
| { |
| "epoch": 32.36, |
| "learning_rate": 0.00013880236185336044, |
| "loss": 0.05, |
| "step": 809 |
| }, |
| { |
| "epoch": 32.4, |
| "learning_rate": 0.00013864991692924523, |
| "loss": 0.0796, |
| "step": 810 |
| }, |
| { |
| "epoch": 32.44, |
| "learning_rate": 0.00013849736633773297, |
| "loss": 0.0617, |
| "step": 811 |
| }, |
| { |
| "epoch": 32.48, |
| "learning_rate": 0.00013834471049589117, |
| "loss": 0.0493, |
| "step": 812 |
| }, |
| { |
| "epoch": 32.52, |
| "learning_rate": 0.00013819194982107503, |
| "loss": 0.0592, |
| "step": 813 |
| }, |
| { |
| "epoch": 32.56, |
| "learning_rate": 0.00013803908473092647, |
| "loss": 0.0663, |
| "step": 814 |
| }, |
| { |
| "epoch": 32.6, |
| "learning_rate": 0.00013788611564337277, |
| "loss": 0.0717, |
| "step": 815 |
| }, |
| { |
| "epoch": 32.64, |
| "learning_rate": 0.00013773304297662559, |
| "loss": 0.0728, |
| "step": 816 |
| }, |
| { |
| "epoch": 32.68, |
| "learning_rate": 0.00013757986714917972, |
| "loss": 0.0574, |
| "step": 817 |
| }, |
| { |
| "epoch": 32.72, |
| "learning_rate": 0.00013742658857981204, |
| "loss": 0.0641, |
| "step": 818 |
| }, |
| { |
| "epoch": 32.76, |
| "learning_rate": 0.00013727320768758033, |
| "loss": 0.0663, |
| "step": 819 |
| }, |
| { |
| "epoch": 32.8, |
| "learning_rate": 0.00013711972489182208, |
| "loss": 0.0598, |
| "step": 820 |
| }, |
| { |
| "epoch": 32.84, |
| "learning_rate": 0.0001369661406121534, |
| "loss": 0.0797, |
| "step": 821 |
| }, |
| { |
| "epoch": 32.88, |
| "learning_rate": 0.00013681245526846783, |
| "loss": 0.0674, |
| "step": 822 |
| }, |
| { |
| "epoch": 32.92, |
| "learning_rate": 0.0001366586692809352, |
| "loss": 0.0766, |
| "step": 823 |
| }, |
| { |
| "epoch": 32.96, |
| "learning_rate": 0.00013650478307000057, |
| "loss": 0.0652, |
| "step": 824 |
| }, |
| { |
| "epoch": 33.0, |
| "learning_rate": 0.00013635079705638298, |
| "loss": 0.0564, |
| "step": 825 |
| }, |
| { |
| "epoch": 33.04, |
| "learning_rate": 0.0001361967116610743, |
| "loss": 0.0466, |
| "step": 826 |
| }, |
| { |
| "epoch": 33.08, |
| "learning_rate": 0.0001360425273053381, |
| "loss": 0.0595, |
| "step": 827 |
| }, |
| { |
| "epoch": 33.12, |
| "learning_rate": 0.00013588824441070852, |
| "loss": 0.0537, |
| "step": 828 |
| }, |
| { |
| "epoch": 33.16, |
| "learning_rate": 0.0001357338633989891, |
| "loss": 0.0642, |
| "step": 829 |
| }, |
| { |
| "epoch": 33.2, |
| "learning_rate": 0.00013557938469225167, |
| "loss": 0.0531, |
| "step": 830 |
| }, |
| { |
| "epoch": 33.24, |
| "learning_rate": 0.00013542480871283507, |
| "loss": 0.0704, |
| "step": 831 |
| }, |
| { |
| "epoch": 33.28, |
| "learning_rate": 0.00013527013588334415, |
| "loss": 0.0458, |
| "step": 832 |
| }, |
| { |
| "epoch": 33.32, |
| "learning_rate": 0.00013511536662664847, |
| "loss": 0.083, |
| "step": 833 |
| }, |
| { |
| "epoch": 33.36, |
| "learning_rate": 0.00013496050136588134, |
| "loss": 0.0745, |
| "step": 834 |
| }, |
| { |
| "epoch": 33.4, |
| "learning_rate": 0.00013480554052443846, |
| "loss": 0.0575, |
| "step": 835 |
| }, |
| { |
| "epoch": 33.44, |
| "learning_rate": 0.00013465048452597682, |
| "loss": 0.0603, |
| "step": 836 |
| }, |
| { |
| "epoch": 33.48, |
| "learning_rate": 0.0001344953337944136, |
| "loss": 0.0766, |
| "step": 837 |
| }, |
| { |
| "epoch": 33.52, |
| "learning_rate": 0.000134340088753925, |
| "loss": 0.0598, |
| "step": 838 |
| }, |
| { |
| "epoch": 33.56, |
| "learning_rate": 0.00013418474982894498, |
| "loss": 0.0664, |
| "step": 839 |
| }, |
| { |
| "epoch": 33.6, |
| "learning_rate": 0.00013402931744416433, |
| "loss": 0.0617, |
| "step": 840 |
| }, |
| { |
| "epoch": 33.64, |
| "learning_rate": 0.00013387379202452917, |
| "loss": 0.0766, |
| "step": 841 |
| }, |
| { |
| "epoch": 33.68, |
| "learning_rate": 0.00013371817399524005, |
| "loss": 0.0754, |
| "step": 842 |
| }, |
| { |
| "epoch": 33.72, |
| "learning_rate": 0.0001335624637817508, |
| "loss": 0.0716, |
| "step": 843 |
| }, |
| { |
| "epoch": 33.76, |
| "learning_rate": 0.00013340666180976712, |
| "loss": 0.0694, |
| "step": 844 |
| }, |
| { |
| "epoch": 33.8, |
| "learning_rate": 0.0001332507685052457, |
| "loss": 0.0574, |
| "step": 845 |
| }, |
| { |
| "epoch": 33.84, |
| "learning_rate": 0.00013309478429439283, |
| "loss": 0.081, |
| "step": 846 |
| }, |
| { |
| "epoch": 33.88, |
| "learning_rate": 0.00013293870960366346, |
| "loss": 0.0717, |
| "step": 847 |
| }, |
| { |
| "epoch": 33.92, |
| "learning_rate": 0.00013278254485975976, |
| "loss": 0.0681, |
| "step": 848 |
| }, |
| { |
| "epoch": 33.96, |
| "learning_rate": 0.00013262629048963026, |
| "loss": 0.0605, |
| "step": 849 |
| }, |
| { |
| "epoch": 34.0, |
| "learning_rate": 0.00013246994692046836, |
| "loss": 0.0447, |
| "step": 850 |
| }, |
| { |
| "epoch": 34.04, |
| "learning_rate": 0.00013231351457971145, |
| "loss": 0.0637, |
| "step": 851 |
| }, |
| { |
| "epoch": 34.08, |
| "learning_rate": 0.00013215699389503954, |
| "loss": 0.0604, |
| "step": 852 |
| }, |
| { |
| "epoch": 34.12, |
| "learning_rate": 0.00013200038529437427, |
| "loss": 0.0723, |
| "step": 853 |
| }, |
| { |
| "epoch": 34.16, |
| "learning_rate": 0.00013184368920587754, |
| "loss": 0.0644, |
| "step": 854 |
| }, |
| { |
| "epoch": 34.2, |
| "learning_rate": 0.00013168690605795045, |
| "loss": 0.0566, |
| "step": 855 |
| }, |
| { |
| "epoch": 34.24, |
| "learning_rate": 0.00013153003627923218, |
| "loss": 0.0591, |
| "step": 856 |
| }, |
| { |
| "epoch": 34.28, |
| "learning_rate": 0.00013137308029859864, |
| "loss": 0.0514, |
| "step": 857 |
| }, |
| { |
| "epoch": 34.32, |
| "learning_rate": 0.0001312160385451616, |
| "loss": 0.0612, |
| "step": 858 |
| }, |
| { |
| "epoch": 34.36, |
| "learning_rate": 0.00013105891144826715, |
| "loss": 0.0815, |
| "step": 859 |
| }, |
| { |
| "epoch": 34.4, |
| "learning_rate": 0.00013090169943749476, |
| "loss": 0.0669, |
| "step": 860 |
| }, |
| { |
| "epoch": 34.44, |
| "learning_rate": 0.0001307444029426561, |
| "loss": 0.0573, |
| "step": 861 |
| }, |
| { |
| "epoch": 34.48, |
| "learning_rate": 0.00013058702239379376, |
| "loss": 0.064, |
| "step": 862 |
| }, |
| { |
| "epoch": 34.52, |
| "learning_rate": 0.00013042955822118016, |
| "loss": 0.0505, |
| "step": 863 |
| }, |
| { |
| "epoch": 34.56, |
| "learning_rate": 0.00013027201085531634, |
| "loss": 0.0714, |
| "step": 864 |
| }, |
| { |
| "epoch": 34.6, |
| "learning_rate": 0.00013011438072693077, |
| "loss": 0.0668, |
| "step": 865 |
| }, |
| { |
| "epoch": 34.64, |
| "learning_rate": 0.00012995666826697819, |
| "loss": 0.0627, |
| "step": 866 |
| }, |
| { |
| "epoch": 34.68, |
| "learning_rate": 0.00012979887390663847, |
| "loss": 0.0469, |
| "step": 867 |
| }, |
| { |
| "epoch": 34.72, |
| "learning_rate": 0.0001296409980773154, |
| "loss": 0.0541, |
| "step": 868 |
| }, |
| { |
| "epoch": 34.76, |
| "learning_rate": 0.00012948304121063536, |
| "loss": 0.0663, |
| "step": 869 |
| }, |
| { |
| "epoch": 34.8, |
| "learning_rate": 0.0001293250037384465, |
| "loss": 0.0721, |
| "step": 870 |
| }, |
| { |
| "epoch": 34.84, |
| "learning_rate": 0.00012916688609281715, |
| "loss": 0.0583, |
| "step": 871 |
| }, |
| { |
| "epoch": 34.88, |
| "learning_rate": 0.00012900868870603503, |
| "loss": 0.0853, |
| "step": 872 |
| }, |
| { |
| "epoch": 34.92, |
| "learning_rate": 0.0001288504120106056, |
| "loss": 0.0528, |
| "step": 873 |
| }, |
| { |
| "epoch": 34.96, |
| "learning_rate": 0.0001286920564392514, |
| "loss": 0.0739, |
| "step": 874 |
| }, |
| { |
| "epoch": 35.0, |
| "learning_rate": 0.00012853362242491053, |
| "loss": 0.0574, |
| "step": 875 |
| }, |
| { |
| "epoch": 35.04, |
| "learning_rate": 0.0001283751104007355, |
| "loss": 0.0581, |
| "step": 876 |
| }, |
| { |
| "epoch": 35.08, |
| "learning_rate": 0.00012821652080009213, |
| "loss": 0.062, |
| "step": 877 |
| }, |
| { |
| "epoch": 35.12, |
| "learning_rate": 0.00012805785405655833, |
| "loss": 0.0492, |
| "step": 878 |
| }, |
| { |
| "epoch": 35.16, |
| "learning_rate": 0.00012789911060392294, |
| "loss": 0.0533, |
| "step": 879 |
| }, |
| { |
| "epoch": 35.2, |
| "learning_rate": 0.00012774029087618446, |
| "loss": 0.0536, |
| "step": 880 |
| }, |
| { |
| "epoch": 35.24, |
| "learning_rate": 0.00012758139530755, |
| "loss": 0.0474, |
| "step": 881 |
| }, |
| { |
| "epoch": 35.28, |
| "learning_rate": 0.00012742242433243396, |
| "loss": 0.0658, |
| "step": 882 |
| }, |
| { |
| "epoch": 35.32, |
| "learning_rate": 0.0001272633783854569, |
| "loss": 0.0816, |
| "step": 883 |
| }, |
| { |
| "epoch": 35.36, |
| "learning_rate": 0.00012710425790144446, |
| "loss": 0.0553, |
| "step": 884 |
| }, |
| { |
| "epoch": 35.4, |
| "learning_rate": 0.0001269450633154258, |
| "loss": 0.0686, |
| "step": 885 |
| }, |
| { |
| "epoch": 35.44, |
| "learning_rate": 0.00012678579506263297, |
| "loss": 0.0816, |
| "step": 886 |
| }, |
| { |
| "epoch": 35.48, |
| "learning_rate": 0.00012662645357849924, |
| "loss": 0.0795, |
| "step": 887 |
| }, |
| { |
| "epoch": 35.52, |
| "learning_rate": 0.00012646703929865817, |
| "loss": 0.0526, |
| "step": 888 |
| }, |
| { |
| "epoch": 35.56, |
| "learning_rate": 0.0001263075526589423, |
| "loss": 0.0548, |
| "step": 889 |
| }, |
| { |
| "epoch": 35.6, |
| "learning_rate": 0.00012614799409538198, |
| "loss": 0.0712, |
| "step": 890 |
| }, |
| { |
| "epoch": 35.64, |
| "learning_rate": 0.0001259883640442043, |
| "loss": 0.0724, |
| "step": 891 |
| }, |
| { |
| "epoch": 35.68, |
| "learning_rate": 0.00012582866294183167, |
| "loss": 0.0666, |
| "step": 892 |
| }, |
| { |
| "epoch": 35.72, |
| "learning_rate": 0.00012566889122488084, |
| "loss": 0.0593, |
| "step": 893 |
| }, |
| { |
| "epoch": 35.76, |
| "learning_rate": 0.00012550904933016153, |
| "loss": 0.0574, |
| "step": 894 |
| }, |
| { |
| "epoch": 35.8, |
| "learning_rate": 0.0001253491376946754, |
| "loss": 0.0504, |
| "step": 895 |
| }, |
| { |
| "epoch": 35.84, |
| "learning_rate": 0.00012518915675561483, |
| "loss": 0.0455, |
| "step": 896 |
| }, |
| { |
| "epoch": 35.88, |
| "learning_rate": 0.00012502910695036146, |
| "loss": 0.0693, |
| "step": 897 |
| }, |
| { |
| "epoch": 35.92, |
| "learning_rate": 0.0001248689887164855, |
| "loss": 0.0643, |
| "step": 898 |
| }, |
| { |
| "epoch": 35.96, |
| "learning_rate": 0.00012470880249174397, |
| "loss": 0.0559, |
| "step": 899 |
| }, |
| { |
| "epoch": 36.0, |
| "learning_rate": 0.00012454854871407994, |
| "loss": 0.0712, |
| "step": 900 |
| }, |
| { |
| "epoch": 36.04, |
| "learning_rate": 0.0001243882278216211, |
| "loss": 0.0423, |
| "step": 901 |
| }, |
| { |
| "epoch": 36.08, |
| "learning_rate": 0.00012422784025267864, |
| "loss": 0.0592, |
| "step": 902 |
| }, |
| { |
| "epoch": 36.12, |
| "learning_rate": 0.00012406738644574612, |
| "loss": 0.0688, |
| "step": 903 |
| }, |
| { |
| "epoch": 36.16, |
| "learning_rate": 0.00012390686683949798, |
| "loss": 0.0609, |
| "step": 904 |
| }, |
| { |
| "epoch": 36.2, |
| "learning_rate": 0.00012374628187278888, |
| "loss": 0.0626, |
| "step": 905 |
| }, |
| { |
| "epoch": 36.24, |
| "learning_rate": 0.00012358563198465182, |
| "loss": 0.0703, |
| "step": 906 |
| }, |
| { |
| "epoch": 36.28, |
| "learning_rate": 0.0001234249176142976, |
| "loss": 0.0651, |
| "step": 907 |
| }, |
| { |
| "epoch": 36.32, |
| "learning_rate": 0.00012326413920111303, |
| "loss": 0.0658, |
| "step": 908 |
| }, |
| { |
| "epoch": 36.36, |
| "learning_rate": 0.00012310329718466024, |
| "loss": 0.0739, |
| "step": 909 |
| }, |
| { |
| "epoch": 36.4, |
| "learning_rate": 0.00012294239200467516, |
| "loss": 0.0621, |
| "step": 910 |
| }, |
| { |
| "epoch": 36.44, |
| "learning_rate": 0.0001227814241010664, |
| "loss": 0.0655, |
| "step": 911 |
| }, |
| { |
| "epoch": 36.48, |
| "learning_rate": 0.00012262039391391404, |
| "loss": 0.077, |
| "step": 912 |
| }, |
| { |
| "epoch": 36.52, |
| "learning_rate": 0.00012245930188346847, |
| "loss": 0.062, |
| "step": 913 |
| }, |
| { |
| "epoch": 36.56, |
| "learning_rate": 0.0001222981484501492, |
| "loss": 0.0605, |
| "step": 914 |
| }, |
| { |
| "epoch": 36.6, |
| "learning_rate": 0.00012213693405454344, |
| "loss": 0.0696, |
| "step": 915 |
| }, |
| { |
| "epoch": 36.64, |
| "learning_rate": 0.00012197565913740531, |
| "loss": 0.0659, |
| "step": 916 |
| }, |
| { |
| "epoch": 36.68, |
| "learning_rate": 0.00012181432413965428, |
| "loss": 0.0547, |
| "step": 917 |
| }, |
| { |
| "epoch": 36.72, |
| "learning_rate": 0.00012165292950237399, |
| "loss": 0.0643, |
| "step": 918 |
| }, |
| { |
| "epoch": 36.76, |
| "learning_rate": 0.00012149147566681124, |
| "loss": 0.0491, |
| "step": 919 |
| }, |
| { |
| "epoch": 36.8, |
| "learning_rate": 0.0001213299630743747, |
| "loss": 0.0637, |
| "step": 920 |
| }, |
| { |
| "epoch": 36.84, |
| "learning_rate": 0.00012116839216663359, |
| "loss": 0.0543, |
| "step": 921 |
| }, |
| { |
| "epoch": 36.88, |
| "learning_rate": 0.0001210067633853166, |
| "loss": 0.0639, |
| "step": 922 |
| }, |
| { |
| "epoch": 36.92, |
| "learning_rate": 0.00012084507717231064, |
| "loss": 0.0605, |
| "step": 923 |
| }, |
| { |
| "epoch": 36.96, |
| "learning_rate": 0.00012068333396965968, |
| "loss": 0.0607, |
| "step": 924 |
| }, |
| { |
| "epoch": 37.0, |
| "learning_rate": 0.00012052153421956342, |
| "loss": 0.0611, |
| "step": 925 |
| }, |
| { |
| "epoch": 37.04, |
| "learning_rate": 0.00012035967836437625, |
| "loss": 0.0561, |
| "step": 926 |
| }, |
| { |
| "epoch": 37.08, |
| "learning_rate": 0.00012019776684660584, |
| "loss": 0.0709, |
| "step": 927 |
| }, |
| { |
| "epoch": 37.12, |
| "learning_rate": 0.00012003580010891213, |
| "loss": 0.0806, |
| "step": 928 |
| }, |
| { |
| "epoch": 37.16, |
| "learning_rate": 0.00011987377859410597, |
| "loss": 0.0639, |
| "step": 929 |
| }, |
| { |
| "epoch": 37.2, |
| "learning_rate": 0.00011971170274514802, |
| "loss": 0.0807, |
| "step": 930 |
| }, |
| { |
| "epoch": 37.24, |
| "learning_rate": 0.00011954957300514748, |
| "loss": 0.059, |
| "step": 931 |
| }, |
| { |
| "epoch": 37.28, |
| "learning_rate": 0.00011938738981736085, |
| "loss": 0.0618, |
| "step": 932 |
| }, |
| { |
| "epoch": 37.32, |
| "learning_rate": 0.0001192251536251907, |
| "loss": 0.0417, |
| "step": 933 |
| }, |
| { |
| "epoch": 37.36, |
| "learning_rate": 0.0001190628648721847, |
| "loss": 0.0715, |
| "step": 934 |
| }, |
| { |
| "epoch": 37.4, |
| "learning_rate": 0.00011890052400203404, |
| "loss": 0.0656, |
| "step": 935 |
| }, |
| { |
| "epoch": 37.44, |
| "learning_rate": 0.00011873813145857249, |
| "loss": 0.0478, |
| "step": 936 |
| }, |
| { |
| "epoch": 37.48, |
| "learning_rate": 0.000118575687685775, |
| "loss": 0.0591, |
| "step": 937 |
| }, |
| { |
| "epoch": 37.52, |
| "learning_rate": 0.00011841319312775671, |
| "loss": 0.0713, |
| "step": 938 |
| }, |
| { |
| "epoch": 37.56, |
| "learning_rate": 0.00011825064822877152, |
| "loss": 0.0574, |
| "step": 939 |
| }, |
| { |
| "epoch": 37.6, |
| "learning_rate": 0.000118088053433211, |
| "loss": 0.0508, |
| "step": 940 |
| }, |
| { |
| "epoch": 37.64, |
| "learning_rate": 0.00011792540918560306, |
| "loss": 0.0685, |
| "step": 941 |
| }, |
| { |
| "epoch": 37.68, |
| "learning_rate": 0.00011776271593061089, |
| "loss": 0.0569, |
| "step": 942 |
| }, |
| { |
| "epoch": 37.72, |
| "learning_rate": 0.00011759997411303161, |
| "loss": 0.0583, |
| "step": 943 |
| }, |
| { |
| "epoch": 37.76, |
| "learning_rate": 0.00011743718417779517, |
| "loss": 0.0636, |
| "step": 944 |
| }, |
| { |
| "epoch": 37.8, |
| "learning_rate": 0.00011727434656996305, |
| "loss": 0.0574, |
| "step": 945 |
| }, |
| { |
| "epoch": 37.84, |
| "learning_rate": 0.000117111461734727, |
| "loss": 0.0586, |
| "step": 946 |
| }, |
| { |
| "epoch": 37.88, |
| "learning_rate": 0.00011694853011740798, |
| "loss": 0.0679, |
| "step": 947 |
| }, |
| { |
| "epoch": 37.92, |
| "learning_rate": 0.00011678555216345477, |
| "loss": 0.0667, |
| "step": 948 |
| }, |
| { |
| "epoch": 37.96, |
| "learning_rate": 0.00011662252831844296, |
| "loss": 0.0716, |
| "step": 949 |
| }, |
| { |
| "epoch": 38.0, |
| "learning_rate": 0.00011645945902807341, |
| "loss": 0.064, |
| "step": 950 |
| }, |
| { |
| "epoch": 38.04, |
| "learning_rate": 0.00011629634473817137, |
| "loss": 0.0622, |
| "step": 951 |
| }, |
| { |
| "epoch": 38.08, |
| "learning_rate": 0.00011613318589468511, |
| "loss": 0.0653, |
| "step": 952 |
| }, |
| { |
| "epoch": 38.12, |
| "learning_rate": 0.00011596998294368466, |
| "loss": 0.0622, |
| "step": 953 |
| }, |
| { |
| "epoch": 38.16, |
| "learning_rate": 0.00011580673633136065, |
| "loss": 0.0546, |
| "step": 954 |
| }, |
| { |
| "epoch": 38.2, |
| "learning_rate": 0.0001156434465040231, |
| "loss": 0.0562, |
| "step": 955 |
| }, |
| { |
| "epoch": 38.24, |
| "learning_rate": 0.00011548011390810017, |
| "loss": 0.0416, |
| "step": 956 |
| }, |
| { |
| "epoch": 38.28, |
| "learning_rate": 0.00011531673899013686, |
| "loss": 0.041, |
| "step": 957 |
| }, |
| { |
| "epoch": 38.32, |
| "learning_rate": 0.00011515332219679404, |
| "loss": 0.0679, |
| "step": 958 |
| }, |
| { |
| "epoch": 38.36, |
| "learning_rate": 0.000114989863974847, |
| "loss": 0.0658, |
| "step": 959 |
| }, |
| { |
| "epoch": 38.4, |
| "learning_rate": 0.0001148263647711842, |
| "loss": 0.0668, |
| "step": 960 |
| }, |
| { |
| "epoch": 38.44, |
| "learning_rate": 0.0001146628250328063, |
| "loss": 0.0662, |
| "step": 961 |
| }, |
| { |
| "epoch": 38.48, |
| "learning_rate": 0.0001144992452068246, |
| "loss": 0.0541, |
| "step": 962 |
| }, |
| { |
| "epoch": 38.52, |
| "learning_rate": 0.00011433562574046019, |
| "loss": 0.0561, |
| "step": 963 |
| }, |
| { |
| "epoch": 38.56, |
| "learning_rate": 0.00011417196708104243, |
| "loss": 0.0651, |
| "step": 964 |
| }, |
| { |
| "epoch": 38.6, |
| "learning_rate": 0.0001140082696760078, |
| "loss": 0.0533, |
| "step": 965 |
| }, |
| { |
| "epoch": 38.64, |
| "learning_rate": 0.00011384453397289876, |
| "loss": 0.059, |
| "step": 966 |
| }, |
| { |
| "epoch": 38.68, |
| "learning_rate": 0.00011368076041936251, |
| "loss": 0.0688, |
| "step": 967 |
| }, |
| { |
| "epoch": 38.72, |
| "learning_rate": 0.0001135169494631497, |
| "loss": 0.0607, |
| "step": 968 |
| }, |
| { |
| "epoch": 38.76, |
| "learning_rate": 0.00011335310155211318, |
| "loss": 0.0705, |
| "step": 969 |
| }, |
| { |
| "epoch": 38.8, |
| "learning_rate": 0.00011318921713420691, |
| "loss": 0.0719, |
| "step": 970 |
| }, |
| { |
| "epoch": 38.84, |
| "learning_rate": 0.0001130252966574846, |
| "loss": 0.081, |
| "step": 971 |
| }, |
| { |
| "epoch": 38.88, |
| "learning_rate": 0.00011286134057009863, |
| "loss": 0.0621, |
| "step": 972 |
| }, |
| { |
| "epoch": 38.92, |
| "learning_rate": 0.00011269734932029863, |
| "loss": 0.0754, |
| "step": 973 |
| }, |
| { |
| "epoch": 38.96, |
| "learning_rate": 0.00011253332335643043, |
| "loss": 0.0685, |
| "step": 974 |
| }, |
| { |
| "epoch": 39.0, |
| "learning_rate": 0.00011236926312693479, |
| "loss": 0.061, |
| "step": 975 |
| }, |
| { |
| "epoch": 39.04, |
| "learning_rate": 0.00011220516908034601, |
| "loss": 0.0532, |
| "step": 976 |
| }, |
| { |
| "epoch": 39.08, |
| "learning_rate": 0.0001120410416652911, |
| "loss": 0.0665, |
| "step": 977 |
| }, |
| { |
| "epoch": 39.12, |
| "learning_rate": 0.00011187688133048801, |
| "loss": 0.0655, |
| "step": 978 |
| }, |
| { |
| "epoch": 39.16, |
| "learning_rate": 0.00011171268852474493, |
| "loss": 0.0665, |
| "step": 979 |
| }, |
| { |
| "epoch": 39.2, |
| "learning_rate": 0.00011154846369695863, |
| "loss": 0.0639, |
| "step": 980 |
| }, |
| { |
| "epoch": 39.24, |
| "learning_rate": 0.00011138420729611355, |
| "loss": 0.0546, |
| "step": 981 |
| }, |
| { |
| "epoch": 39.28, |
| "learning_rate": 0.00011121991977128045, |
| "loss": 0.0724, |
| "step": 982 |
| }, |
| { |
| "epoch": 39.32, |
| "learning_rate": 0.00011105560157161513, |
| "loss": 0.064, |
| "step": 983 |
| }, |
| { |
| "epoch": 39.36, |
| "learning_rate": 0.00011089125314635726, |
| "loss": 0.07, |
| "step": 984 |
| }, |
| { |
| "epoch": 39.4, |
| "learning_rate": 0.00011072687494482919, |
| "loss": 0.0522, |
| "step": 985 |
| }, |
| { |
| "epoch": 39.44, |
| "learning_rate": 0.0001105624674164346, |
| "loss": 0.0463, |
| "step": 986 |
| }, |
| { |
| "epoch": 39.48, |
| "learning_rate": 0.00011039803101065742, |
| "loss": 0.055, |
| "step": 987 |
| }, |
| { |
| "epoch": 39.52, |
| "learning_rate": 0.00011023356617706052, |
| "loss": 0.0664, |
| "step": 988 |
| }, |
| { |
| "epoch": 39.56, |
| "learning_rate": 0.00011006907336528445, |
| "loss": 0.0577, |
| "step": 989 |
| }, |
| { |
| "epoch": 39.6, |
| "learning_rate": 0.0001099045530250463, |
| "loss": 0.0591, |
| "step": 990 |
| }, |
| { |
| "epoch": 39.64, |
| "learning_rate": 0.00010974000560613835, |
| "loss": 0.0656, |
| "step": 991 |
| }, |
| { |
| "epoch": 39.68, |
| "learning_rate": 0.00010957543155842702, |
| "loss": 0.065, |
| "step": 992 |
| }, |
| { |
| "epoch": 39.72, |
| "learning_rate": 0.00010941083133185146, |
| "loss": 0.0683, |
| "step": 993 |
| }, |
| { |
| "epoch": 39.76, |
| "learning_rate": 0.00010924620537642236, |
| "loss": 0.0489, |
| "step": 994 |
| }, |
| { |
| "epoch": 39.8, |
| "learning_rate": 0.00010908155414222083, |
| "loss": 0.0552, |
| "step": 995 |
| }, |
| { |
| "epoch": 39.84, |
| "learning_rate": 0.00010891687807939707, |
| "loss": 0.0676, |
| "step": 996 |
| }, |
| { |
| "epoch": 39.88, |
| "learning_rate": 0.00010875217763816911, |
| "loss": 0.068, |
| "step": 997 |
| }, |
| { |
| "epoch": 39.92, |
| "learning_rate": 0.00010858745326882171, |
| "loss": 0.0737, |
| "step": 998 |
| }, |
| { |
| "epoch": 39.96, |
| "learning_rate": 0.00010842270542170499, |
| "loss": 0.0639, |
| "step": 999 |
| }, |
| { |
| "epoch": 40.0, |
| "learning_rate": 0.00010825793454723325, |
| "loss": 0.0604, |
| "step": 1000 |
| } |
| ], |
| "max_steps": 2000, |
| "num_train_epochs": 80, |
| "total_flos": 7047420037632000.0, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
