dzungpham/graphcodebert-code-classification
0
1{2 "best_global_step": null,3 "best_metric": null,4 "best_model_checkpoint": null,5 "epoch": 0.538160469667319,6 "eval_steps": 1000,7 "global_step": 550,8 "is_hyper_param_search": false,9 "is_local_process_zero": true,10 "is_world_process_zero": true,11 "log_history": [12 {13 "epoch": 0.009784735812133072,14 "grad_norm": 165951.1875,15 "learning_rate": 4.411764705882353e-08,16 "loss": 0.8416,17 "step": 1018 },19 {20 "epoch": 0.019569471624266144,21 "grad_norm": 206132.796875,22 "learning_rate": 9.313725490196079e-08,23 "loss": 0.8378,24 "step": 2025 },26 {27 "epoch": 0.029354207436399216,28 "grad_norm": 171197.53125,29 "learning_rate": 1.4215686274509803e-07,30 "loss": 0.8388,31 "step": 3032 },33 {34 "epoch": 0.03913894324853229,35 "grad_norm": 206661.828125,36 "learning_rate": 1.9117647058823527e-07,37 "loss": 0.847,38 "step": 4039 },40 {41 "epoch": 0.04892367906066536,42 "grad_norm": 230268.96875,43 "learning_rate": 2.4019607843137255e-07,44 "loss": 0.8411,45 "step": 5046 },47 {48 "epoch": 0.05870841487279843,49 "grad_norm": 133248.28125,50 "learning_rate": 2.8921568627450984e-07,51 "loss": 0.8412,52 "step": 6053 },54 {55 "epoch": 0.0684931506849315,56 "grad_norm": 147768.828125,57 "learning_rate": 3.3823529411764707e-07,58 "loss": 0.8431,59 "step": 7060 },61 {62 "epoch": 0.07827788649706457,63 "grad_norm": 167327.53125,64 "learning_rate": 3.872549019607843e-07,65 "loss": 0.8431,66 "step": 8067 },68 {69 "epoch": 0.08806262230919765,70 "grad_norm": 151943.75,71 "learning_rate": 4.3627450980392154e-07,72 "loss": 0.8311,73 "step": 9074 },75 {76 "epoch": 0.09784735812133072,77 "grad_norm": 149971.59375,78 "learning_rate": 4.852941176470588e-07,79 "loss": 0.8412,80 "step": 10081 },82 {83 "epoch": 0.10763209393346379,84 "grad_norm": 162329.40625,85 "learning_rate": 5.34313725490196e-07,86 "loss": 0.8356,87 "step": 11088 },89 {90 "epoch": 0.11741682974559686,91 "grad_norm": 158769.765625,92 "learning_rate": 5.833333333333334e-07,93 "loss": 0.8351,94 "step": 12095 },96 {97 "epoch": 0.12720156555772993,98 "grad_norm": 157725.59375,99 "learning_rate": 6.323529411764706e-07,100 "loss": 0.8331,101 "step": 130102 },103 {104 "epoch": 0.136986301369863,105 "grad_norm": 205733.0,106 "learning_rate": 6.813725490196079e-07,107 "loss": 0.8366,108 "step": 140109 },110 {111 "epoch": 0.14677103718199608,112 "grad_norm": 172581.328125,113 "learning_rate": 7.30392156862745e-07,114 "loss": 0.831,115 "step": 150116 },117 {118 "epoch": 0.15655577299412915,119 "grad_norm": 137428.0,120 "learning_rate": 7.794117647058823e-07,121 "loss": 0.8326,122 "step": 160123 },124 {125 "epoch": 0.16634050880626222,126 "grad_norm": 151189.203125,127 "learning_rate": 8.284313725490196e-07,128 "loss": 0.8281,129 "step": 170130 },131 {132 "epoch": 0.1761252446183953,133 "grad_norm": 145620.625,134 "learning_rate": 8.774509803921568e-07,135 "loss": 0.8252,136 "step": 180137 },138 {139 "epoch": 0.18590998043052837,140 "grad_norm": 169938.84375,141 "learning_rate": 9.264705882352941e-07,142 "loss": 0.8308,143 "step": 190144 },145 {146 "epoch": 0.19569471624266144,147 "grad_norm": 138760.015625,148 "learning_rate": 9.754901960784313e-07,149 "loss": 0.824,150 "step": 200151 },152 {153 "epoch": 0.2054794520547945,154 "grad_norm": 107872.125,155 "learning_rate": 9.99631294374483e-07,156 "loss": 0.8254,157 "step": 210158 },159 {160 "epoch": 0.21526418786692758,161 "grad_norm": 113610.8359375,162 "learning_rate": 9.966849112204936e-07,163 "loss": 0.8267,164 "step": 220165 },166 {167 "epoch": 0.22504892367906065,168 "grad_norm": 172441.515625,169 "learning_rate": 9.908095200725339e-07,170 "loss": 0.8198,171 "step": 230172 },173 {174 "epoch": 0.23483365949119372,175 "grad_norm": 128079.9140625,176 "learning_rate": 9.82039768787326e-07,177 "loss": 0.8143,178 "step": 240179 },180 {181 "epoch": 0.2446183953033268,182 "grad_norm": 98149.8984375,183 "learning_rate": 9.70427373595903e-07,184 "loss": 0.8191,185 "step": 250186 },187 {188 "epoch": 0.25440313111545987,189 "grad_norm": 147377.4375,190 "learning_rate": 9.560408141270531e-07,191 "loss": 0.8178,192 "step": 260193 },194 {195 "epoch": 0.26418786692759294,196 "grad_norm": 156439.40625,197 "learning_rate": 9.389649295750781e-07,198 "loss": 0.8085,199 "step": 270200 },201 {202 "epoch": 0.273972602739726,203 "grad_norm": 131892.015625,204 "learning_rate": 9.193004183933174e-07,205 "loss": 0.8172,206 "step": 280207 },208 {209 "epoch": 0.2837573385518591,210 "grad_norm": 132474.5,211 "learning_rate": 8.971632444637971e-07,212 "loss": 0.8136,213 "step": 290214 },215 {216 "epoch": 0.29354207436399216,217 "grad_norm": 112619.109375,218 "learning_rate": 8.726839532448928e-07,219 "loss": 0.8065,220 "step": 300221 },222 {223 "epoch": 0.30332681017612523,224 "grad_norm": 94225.5703125,225 "learning_rate": 8.460069019297606e-07,226 "loss": 0.813,227 "step": 310228 },229 {230 "epoch": 0.3131115459882583,231 "grad_norm": 117812.8203125,232 "learning_rate": 8.172894081553822e-07,233 "loss": 0.8022,234 "step": 320235 },236 {237 "epoch": 0.32289628180039137,238 "grad_norm": 143287.125,239 "learning_rate": 7.86700822282382e-07,240 "loss": 0.8061,241 "step": 330242 },243 {244 "epoch": 0.33268101761252444,245 "grad_norm": 107456.25,246 "learning_rate": 7.544215287164917e-07,247 "loss": 0.8006,248 "step": 340249 },250 {251 "epoch": 0.3424657534246575,252 "grad_norm": 101521.171875,253 "learning_rate": 7.206418821609812e-07,254 "loss": 0.8055,255 "step": 350256 },257 {258 "epoch": 0.3522504892367906,259 "grad_norm": 118311.5234375,260 "learning_rate": 6.855610850731017e-07,261 "loss": 0.8081,262 "step": 360263 },264 {265 "epoch": 0.36203522504892366,266 "grad_norm": 89621.828125,267 "learning_rate": 6.493860129443046e-07,268 "loss": 0.8058,269 "step": 370270 },271 {272 "epoch": 0.37181996086105673,273 "grad_norm": 143168.65625,274 "learning_rate": 6.123299943316975e-07,275 "loss": 0.8028,276 "step": 380277 },278 {279 "epoch": 0.3816046966731898,280 "grad_norm": 118598.2890625,281 "learning_rate": 5.746115528350295e-07,282 "loss": 0.8032,283 "step": 390284 },285 {286 "epoch": 0.3913894324853229,287 "grad_norm": 111807.8203125,288 "learning_rate": 5.364531184379139e-07,289 "loss": 0.8014,290 "step": 400291 },292 {293 "epoch": 0.40117416829745595,294 "grad_norm": 124106.1328125,295 "learning_rate": 4.980797158126566e-07,296 "loss": 0.8045,297 "step": 410298 },299 {300 "epoch": 0.410958904109589,301 "grad_norm": 102746.4765625,302 "learning_rate": 4.597176373239061e-07,303 "loss": 0.7996,304 "step": 420305 },306 {307 "epoch": 0.4207436399217221,308 "grad_norm": 108317.5859375,309 "learning_rate": 4.215931085565761e-07,310 "loss": 0.8037,311 "step": 430312 },313 {314 "epoch": 0.43052837573385516,315 "grad_norm": 113342.515625,316 "learning_rate": 3.8393095423757764e-07,317 "loss": 0.8047,318 "step": 440319 },320 {321 "epoch": 0.44031311154598823,322 "grad_norm": 111908.90625,323 "learning_rate": 3.469532724185699e-07,324 "loss": 0.8021,325 "step": 450326 },327 {328 "epoch": 0.4500978473581213,329 "grad_norm": 85004.15625,330 "learning_rate": 3.1087812473822845e-07,331 "loss": 0.8012,332 "step": 460333 },334 {335 "epoch": 0.4598825831702544,336 "grad_norm": 98335.1796875,337 "learning_rate": 2.7591825048770643e-07,338 "loss": 0.7977,339 "step": 470340 },341 {342 "epoch": 0.46966731898238745,343 "grad_norm": 110674.03125,344 "learning_rate": 2.422798120625886e-07,345 "loss": 0.7979,346 "step": 480347 },348 {349 "epoch": 0.4794520547945205,350 "grad_norm": 114861.5390625,351 "learning_rate": 2.1016117919955512e-07,352 "loss": 0.8014,353 "step": 490354 },355 {356 "epoch": 0.4892367906066536,357 "grad_norm": 130109.5703125,358 "learning_rate": 1.7975175916725031e-07,359 "loss": 0.7994,360 "step": 500361 },362 {363 "epoch": 0.49902152641878667,364 "grad_norm": 100678.8671875,365 "learning_rate": 1.5123087980985606e-07,366 "loss": 0.802,367 "step": 510368 },369 {370 "epoch": 0.5088062622309197,371 "grad_norm": 115718.078125,372 "learning_rate": 1.2476673203018728e-07,373 "loss": 0.7971,374 "step": 520375 },376 {377 "epoch": 0.5185909980430529,378 "grad_norm": 94845.9453125,379 "learning_rate": 1.0051537794861287e-07,380 "loss": 0.7952,381 "step": 530382 },383 {384 "epoch": 0.5283757338551859,385 "grad_norm": 106283.5703125,386 "learning_rate": 7.861983058679871e-08,387 "loss": 0.8013,388 "step": 540389 },390 {391 "epoch": 0.538160469667319,392 "grad_norm": 93847.1640625,393 "learning_rate": 5.920921050348626e-08,394 "loss": 0.7968,395 "step": 550396 }397 ],398 "logging_steps": 10,399 "max_steps": 1022,400 "num_input_tokens_seen": 0,401 "num_train_epochs": 1,402 "save_steps": 50,403 "stateful_callbacks": {404 "EarlyStoppingCallback": {405 "args": {406 "early_stopping_patience": 3,407 "early_stopping_threshold": 0.0408 },409 "attributes": {410 "early_stopping_patience_counter": 0411 }412 },413 "TrainerControl": {414 "args": {415 "should_epoch_stop": false,416 "should_evaluate": false,417 "should_log": false,418 "should_save": true,419 "should_training_stop": false420 },421 "attributes": {}422 }423 },424 "total_flos": 7.4092073189376e+16,425 "train_batch_size": 512,426 "trial_name": null,427 "trial_params": null428}429