Commit
·
7a9132b
1
Parent(s):
52a09a5
Delete trainer_state.json
Browse files- trainer_state.json +0 -212
trainer_state.json
DELETED
|
@@ -1,212 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"best_metric": 0.8479087452471483,
|
| 3 |
-
"best_model_checkpoint": "./output/model_1\\run-0\\checkpoint-740",
|
| 4 |
-
"epoch": 14.0,
|
| 5 |
-
"global_step": 2072,
|
| 6 |
-
"is_hyper_param_search": true,
|
| 7 |
-
"is_local_process_zero": true,
|
| 8 |
-
"is_world_process_zero": true,
|
| 9 |
-
"log_history": [
|
| 10 |
-
{
|
| 11 |
-
"epoch": 1.0,
|
| 12 |
-
"eval_accuracy": 0.8060836501901141,
|
| 13 |
-
"eval_f1": 0.8000895647160824,
|
| 14 |
-
"eval_loss": 0.46571293473243713,
|
| 15 |
-
"eval_precision": 0.8103755012570384,
|
| 16 |
-
"eval_recall": 0.8060836501901141,
|
| 17 |
-
"eval_runtime": 0.224,
|
| 18 |
-
"eval_samples_per_second": 1174.099,
|
| 19 |
-
"eval_steps_per_second": 75.892,
|
| 20 |
-
"step": 148
|
| 21 |
-
},
|
| 22 |
-
{
|
| 23 |
-
"epoch": 2.0,
|
| 24 |
-
"eval_accuracy": 0.8174904942965779,
|
| 25 |
-
"eval_f1": 0.8174904942965779,
|
| 26 |
-
"eval_loss": 0.49181994795799255,
|
| 27 |
-
"eval_precision": 0.8174904942965779,
|
| 28 |
-
"eval_recall": 0.8174904942965779,
|
| 29 |
-
"eval_runtime": 0.2151,
|
| 30 |
-
"eval_samples_per_second": 1222.875,
|
| 31 |
-
"eval_steps_per_second": 79.045,
|
| 32 |
-
"step": 296
|
| 33 |
-
},
|
| 34 |
-
{
|
| 35 |
-
"epoch": 3.0,
|
| 36 |
-
"eval_accuracy": 0.8174904942965779,
|
| 37 |
-
"eval_f1": 0.8168592417586474,
|
| 38 |
-
"eval_loss": 0.4677908420562744,
|
| 39 |
-
"eval_precision": 0.8166185862354719,
|
| 40 |
-
"eval_recall": 0.8174904942965779,
|
| 41 |
-
"eval_runtime": 0.212,
|
| 42 |
-
"eval_samples_per_second": 1240.567,
|
| 43 |
-
"eval_steps_per_second": 80.189,
|
| 44 |
-
"step": 444
|
| 45 |
-
},
|
| 46 |
-
{
|
| 47 |
-
"epoch": 3.38,
|
| 48 |
-
"learning_rate": 5.031969358167588e-06,
|
| 49 |
-
"loss": 0.4058,
|
| 50 |
-
"step": 500
|
| 51 |
-
},
|
| 52 |
-
{
|
| 53 |
-
"epoch": 4.0,
|
| 54 |
-
"eval_accuracy": 0.8403041825095057,
|
| 55 |
-
"eval_f1": 0.8397518365388166,
|
| 56 |
-
"eval_loss": 0.4895302653312683,
|
| 57 |
-
"eval_precision": 0.8396023801788713,
|
| 58 |
-
"eval_recall": 0.8403041825095057,
|
| 59 |
-
"eval_runtime": 0.2159,
|
| 60 |
-
"eval_samples_per_second": 1218.057,
|
| 61 |
-
"eval_steps_per_second": 78.734,
|
| 62 |
-
"step": 592
|
| 63 |
-
},
|
| 64 |
-
{
|
| 65 |
-
"epoch": 5.0,
|
| 66 |
-
"eval_accuracy": 0.8479087452471483,
|
| 67 |
-
"eval_f1": 0.8470891084157254,
|
| 68 |
-
"eval_loss": 0.5246260762214661,
|
| 69 |
-
"eval_precision": 0.8471907238947951,
|
| 70 |
-
"eval_recall": 0.8479087452471483,
|
| 71 |
-
"eval_runtime": 0.212,
|
| 72 |
-
"eval_samples_per_second": 1240.567,
|
| 73 |
-
"eval_steps_per_second": 80.189,
|
| 74 |
-
"step": 740
|
| 75 |
-
},
|
| 76 |
-
{
|
| 77 |
-
"epoch": 6.0,
|
| 78 |
-
"eval_accuracy": 0.844106463878327,
|
| 79 |
-
"eval_f1": 0.842419911092538,
|
| 80 |
-
"eval_loss": 0.5946373343467712,
|
| 81 |
-
"eval_precision": 0.8438668765907662,
|
| 82 |
-
"eval_recall": 0.844106463878327,
|
| 83 |
-
"eval_runtime": 0.2206,
|
| 84 |
-
"eval_samples_per_second": 1192.052,
|
| 85 |
-
"eval_steps_per_second": 77.053,
|
| 86 |
-
"step": 888
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"epoch": 6.76,
|
| 90 |
-
"learning_rate": 3.4314701984450724e-06,
|
| 91 |
-
"loss": 0.2291,
|
| 92 |
-
"step": 1000
|
| 93 |
-
},
|
| 94 |
-
{
|
| 95 |
-
"epoch": 7.0,
|
| 96 |
-
"eval_accuracy": 0.844106463878327,
|
| 97 |
-
"eval_f1": 0.8437098422789988,
|
| 98 |
-
"eval_loss": 0.650194525718689,
|
| 99 |
-
"eval_precision": 0.8435403587786577,
|
| 100 |
-
"eval_recall": 0.844106463878327,
|
| 101 |
-
"eval_runtime": 0.2125,
|
| 102 |
-
"eval_samples_per_second": 1237.62,
|
| 103 |
-
"eval_steps_per_second": 79.998,
|
| 104 |
-
"step": 1036
|
| 105 |
-
},
|
| 106 |
-
{
|
| 107 |
-
"epoch": 8.0,
|
| 108 |
-
"eval_accuracy": 0.8403041825095057,
|
| 109 |
-
"eval_f1": 0.8391134423962042,
|
| 110 |
-
"eval_loss": 0.6312099099159241,
|
| 111 |
-
"eval_precision": 0.8395542107279025,
|
| 112 |
-
"eval_recall": 0.8403041825095057,
|
| 113 |
-
"eval_runtime": 0.2165,
|
| 114 |
-
"eval_samples_per_second": 1214.746,
|
| 115 |
-
"eval_steps_per_second": 78.52,
|
| 116 |
-
"step": 1184
|
| 117 |
-
},
|
| 118 |
-
{
|
| 119 |
-
"epoch": 9.0,
|
| 120 |
-
"eval_accuracy": 0.8365019011406845,
|
| 121 |
-
"eval_f1": 0.8351053157727012,
|
| 122 |
-
"eval_loss": 0.7456679344177246,
|
| 123 |
-
"eval_precision": 0.8357697627558571,
|
| 124 |
-
"eval_recall": 0.8365019011406845,
|
| 125 |
-
"eval_runtime": 0.2137,
|
| 126 |
-
"eval_samples_per_second": 1230.687,
|
| 127 |
-
"eval_steps_per_second": 79.55,
|
| 128 |
-
"step": 1332
|
| 129 |
-
},
|
| 130 |
-
{
|
| 131 |
-
"epoch": 10.0,
|
| 132 |
-
"eval_accuracy": 0.8403041825095057,
|
| 133 |
-
"eval_f1": 0.838761109426392,
|
| 134 |
-
"eval_loss": 0.7921389937400818,
|
| 135 |
-
"eval_precision": 0.8397848232749173,
|
| 136 |
-
"eval_recall": 0.8403041825095057,
|
| 137 |
-
"eval_runtime": 0.2155,
|
| 138 |
-
"eval_samples_per_second": 1220.389,
|
| 139 |
-
"eval_steps_per_second": 78.884,
|
| 140 |
-
"step": 1480
|
| 141 |
-
},
|
| 142 |
-
{
|
| 143 |
-
"epoch": 10.14,
|
| 144 |
-
"learning_rate": 1.8309710387225575e-06,
|
| 145 |
-
"loss": 0.1587,
|
| 146 |
-
"step": 1500
|
| 147 |
-
},
|
| 148 |
-
{
|
| 149 |
-
"epoch": 11.0,
|
| 150 |
-
"eval_accuracy": 0.8250950570342205,
|
| 151 |
-
"eval_f1": 0.8256084971375828,
|
| 152 |
-
"eval_loss": 0.8698846697807312,
|
| 153 |
-
"eval_precision": 0.8265094169095286,
|
| 154 |
-
"eval_recall": 0.8250950570342205,
|
| 155 |
-
"eval_runtime": 0.219,
|
| 156 |
-
"eval_samples_per_second": 1200.921,
|
| 157 |
-
"eval_steps_per_second": 77.626,
|
| 158 |
-
"step": 1628
|
| 159 |
-
},
|
| 160 |
-
{
|
| 161 |
-
"epoch": 12.0,
|
| 162 |
-
"eval_accuracy": 0.8326996197718631,
|
| 163 |
-
"eval_f1": 0.831083067018125,
|
| 164 |
-
"eval_loss": 0.9105260372161865,
|
| 165 |
-
"eval_precision": 0.832006346665142,
|
| 166 |
-
"eval_recall": 0.8326996197718631,
|
| 167 |
-
"eval_runtime": 0.2169,
|
| 168 |
-
"eval_samples_per_second": 1212.315,
|
| 169 |
-
"eval_steps_per_second": 78.363,
|
| 170 |
-
"step": 1776
|
| 171 |
-
},
|
| 172 |
-
{
|
| 173 |
-
"epoch": 13.0,
|
| 174 |
-
"eval_accuracy": 0.8365019011406845,
|
| 175 |
-
"eval_f1": 0.8354546196277604,
|
| 176 |
-
"eval_loss": 0.9163150787353516,
|
| 177 |
-
"eval_precision": 0.8356459763303871,
|
| 178 |
-
"eval_recall": 0.8365019011406845,
|
| 179 |
-
"eval_runtime": 0.2143,
|
| 180 |
-
"eval_samples_per_second": 1227.36,
|
| 181 |
-
"eval_steps_per_second": 79.335,
|
| 182 |
-
"step": 1924
|
| 183 |
-
},
|
| 184 |
-
{
|
| 185 |
-
"epoch": 13.51,
|
| 186 |
-
"learning_rate": 2.304718790000422e-07,
|
| 187 |
-
"loss": 0.1181,
|
| 188 |
-
"step": 2000
|
| 189 |
-
},
|
| 190 |
-
{
|
| 191 |
-
"epoch": 14.0,
|
| 192 |
-
"eval_accuracy": 0.8326996197718631,
|
| 193 |
-
"eval_f1": 0.8314521777484043,
|
| 194 |
-
"eval_loss": 0.931361734867096,
|
| 195 |
-
"eval_precision": 0.8318193002498804,
|
| 196 |
-
"eval_recall": 0.8326996197718631,
|
| 197 |
-
"eval_runtime": 0.2146,
|
| 198 |
-
"eval_samples_per_second": 1225.552,
|
| 199 |
-
"eval_steps_per_second": 79.218,
|
| 200 |
-
"step": 2072
|
| 201 |
-
}
|
| 202 |
-
],
|
| 203 |
-
"max_steps": 2072,
|
| 204 |
-
"num_train_epochs": 14,
|
| 205 |
-
"total_flos": 131021083348800.0,
|
| 206 |
-
"trial_name": null,
|
| 207 |
-
"trial_params": {
|
| 208 |
-
"learning_rate": 6.632468517890103e-06,
|
| 209 |
-
"num_train_epochs": 14,
|
| 210 |
-
"weight_decay": 0.0028145972556210797
|
| 211 |
-
}
|
| 212 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|