dzungpham/graphcodebert-code-classification

1{2  "best_global_step": null,3  "best_metric": null,4  "best_model_checkpoint": null,5  "epoch": 0.0256,6  "eval_steps": 1000,7  "global_step": 400,8  "is_hyper_param_search": false,9  "is_local_process_zero": true,10  "is_world_process_zero": true,11  "log_history": [12    {13      "epoch": 0.00064,14      "grad_norm": 1.8688431978225708,15      "learning_rate": 3.840409643695328e-08,16      "loss": 0.7168,17      "step": 1018    },19    {20      "epoch": 0.00128,21      "grad_norm": 1.835353970527649,22      "learning_rate": 8.10753147002347e-08,23      "loss": 0.7179,24      "step": 2025    },26    {27      "epoch": 0.00192,28      "grad_norm": 1.2541388273239136,29      "learning_rate": 1.2374653296351612e-07,30      "loss": 0.7177,31      "step": 3032    },33    {34      "epoch": 0.00256,35      "grad_norm": 1.1381633281707764,36      "learning_rate": 1.6641775122679754e-07,37      "loss": 0.7312,38      "step": 4039    },40    {41      "epoch": 0.0032,42      "grad_norm": 1.6313824653625488,43      "learning_rate": 2.0908896949007894e-07,44      "loss": 0.7167,45      "step": 5046    },47    {48      "epoch": 0.00384,49      "grad_norm": 1.092464566230774,50      "learning_rate": 2.517601877533604e-07,51      "loss": 0.7201,52      "step": 6053    },54    {55      "epoch": 0.00448,56      "grad_norm": 1.164099097251892,57      "learning_rate": 2.944314060166418e-07,58      "loss": 0.7195,59      "step": 7060    },61    {62      "epoch": 0.00512,63      "grad_norm": 1.9283920526504517,64      "learning_rate": 3.371026242799232e-07,65      "loss": 0.7223,66      "step": 8067    },68    {69      "epoch": 0.00576,70      "grad_norm": 1.8716129064559937,71      "learning_rate": 3.7977384254320464e-07,72      "loss": 0.7151,73      "step": 9074    },75    {76      "epoch": 0.0064,77      "grad_norm": 1.5499262809753418,78      "learning_rate": 4.22445060806486e-07,79      "loss": 0.7211,80      "step": 10081    },82    {83      "epoch": 0.00704,84      "grad_norm": 1.375114917755127,85      "learning_rate": 4.651162790697675e-07,86      "loss": 0.7247,87      "step": 11088    },89    {90      "epoch": 0.00768,91      "grad_norm": 1.384915828704834,92      "learning_rate": 5.077874973330489e-07,93      "loss": 0.7156,94      "step": 12095    },96    {97      "epoch": 0.00832,98      "grad_norm": 1.6376659870147705,99      "learning_rate": 5.504587155963304e-07,100      "loss": 0.7196,101      "step": 130102    },103    {104      "epoch": 0.00896,105      "grad_norm": 1.709489345550537,106      "learning_rate": 5.931299338596117e-07,107      "loss": 0.7215,108      "step": 140109    },110    {111      "epoch": 0.0096,112      "grad_norm": 1.350512146949768,113      "learning_rate": 6.358011521228932e-07,114      "loss": 0.71,115      "step": 150116    },117    {118      "epoch": 0.01024,119      "grad_norm": 2.33050537109375,120      "learning_rate": 6.784723703861745e-07,121      "loss": 0.7191,122      "step": 160123    },124    {125      "epoch": 0.01088,126      "grad_norm": 1.0042874813079834,127      "learning_rate": 7.21143588649456e-07,128      "loss": 0.72,129      "step": 170130    },131    {132      "epoch": 0.01152,133      "grad_norm": 1.1835744380950928,134      "learning_rate": 7.638148069127374e-07,135      "loss": 0.7122,136      "step": 180137    },138    {139      "epoch": 0.01216,140      "grad_norm": 1.949506402015686,141      "learning_rate": 8.064860251760189e-07,142      "loss": 0.7091,143      "step": 190144    },145    {146      "epoch": 0.0128,147      "grad_norm": 1.139626383781433,148      "learning_rate": 8.491572434393003e-07,149      "loss": 0.7046,150      "step": 200151    },152    {153      "epoch": 0.01344,154      "grad_norm": 1.7734779119491577,155      "learning_rate": 8.918284617025817e-07,156      "loss": 0.7135,157      "step": 210158    },159    {160      "epoch": 0.01408,161      "grad_norm": 1.3442974090576172,162      "learning_rate": 9.344996799658632e-07,163      "loss": 0.7127,164      "step": 220165    },166    {167      "epoch": 0.01472,168      "grad_norm": 1.6148335933685303,169      "learning_rate": 9.771708982291445e-07,170      "loss": 0.7168,171      "step": 230172    },173    {174      "epoch": 0.01536,175      "grad_norm": 0.9220213294029236,176      "learning_rate": 1.0198421164924258e-06,177      "loss": 0.7109,178      "step": 240179    },180    {181      "epoch": 0.016,182      "grad_norm": 1.5631815195083618,183      "learning_rate": 1.0625133347557074e-06,184      "loss": 0.7149,185      "step": 250186    },187    {188      "epoch": 0.01664,189      "grad_norm": 1.5052822828292847,190      "learning_rate": 1.1051845530189888e-06,191      "loss": 0.7112,192      "step": 260193    },194    {195      "epoch": 0.01728,196      "grad_norm": 1.6714903116226196,197      "learning_rate": 1.1478557712822702e-06,198      "loss": 0.7128,199      "step": 270200    },201    {202      "epoch": 0.01792,203      "grad_norm": 1.6667262315750122,204      "learning_rate": 1.1905269895455517e-06,205      "loss": 0.7091,206      "step": 280207    },208    {209      "epoch": 0.01856,210      "grad_norm": 1.6861907243728638,211      "learning_rate": 1.233198207808833e-06,212      "loss": 0.7078,213      "step": 290214    },215    {216      "epoch": 0.0192,217      "grad_norm": 1.6259212493896484,218      "learning_rate": 1.2758694260721145e-06,219      "loss": 0.7087,220      "step": 300221    },222    {223      "epoch": 0.01984,224      "grad_norm": 1.207320213317871,225      "learning_rate": 1.318540644335396e-06,226      "loss": 0.7087,227      "step": 310228    },229    {230      "epoch": 0.02048,231      "grad_norm": 1.087124228477478,232      "learning_rate": 1.3612118625986772e-06,233      "loss": 0.7174,234      "step": 320235    },236    {237      "epoch": 0.02112,238      "grad_norm": 1.5668721199035645,239      "learning_rate": 1.4038830808619588e-06,240      "loss": 0.7066,241      "step": 330242    },243    {244      "epoch": 0.02176,245      "grad_norm": 1.5332859754562378,246      "learning_rate": 1.4465542991252401e-06,247      "loss": 0.7115,248      "step": 340249    },250    {251      "epoch": 0.0224,252      "grad_norm": 2.092994451522827,253      "learning_rate": 1.4892255173885215e-06,254      "loss": 0.7085,255      "step": 350256    },257    {258      "epoch": 0.02304,259      "grad_norm": 1.266851544380188,260      "learning_rate": 1.531896735651803e-06,261      "loss": 0.7091,262      "step": 360263    },264    {265      "epoch": 0.02368,266      "grad_norm": 1.290616512298584,267      "learning_rate": 1.5745679539150842e-06,268      "loss": 0.7103,269      "step": 370270    },271    {272      "epoch": 0.02432,273      "grad_norm": 1.2045000791549683,274      "learning_rate": 1.6172391721783658e-06,275      "loss": 0.705,276      "step": 380277    },278    {279      "epoch": 0.02496,280      "grad_norm": 1.828832983970642,281      "learning_rate": 1.6599103904416472e-06,282      "loss": 0.7038,283      "step": 390284    },285    {286      "epoch": 0.0256,287      "grad_norm": 1.7793196439743042,288      "learning_rate": 1.7025816087049288e-06,289      "loss": 0.7085,290      "step": 400291    }292  ],293  "logging_steps": 10,294  "max_steps": 46875,295  "num_input_tokens_seen": 0,296  "num_train_epochs": 3,297  "save_steps": 200,298  "stateful_callbacks": {299    "EarlyStoppingCallback": {300      "args": {301        "early_stopping_patience": 3,302        "early_stopping_threshold": 0.0303      },304      "attributes": {305        "early_stopping_patience_counter": 0306      }307    },308    "TrainerControl": {309      "args": {310        "should_epoch_stop": false,311        "should_evaluate": false,312        "should_log": false,313        "should_save": true,314        "should_training_stop": false315      },316      "attributes": {}317    }318  },319  "total_flos": 3366473064449280.0,320  "train_batch_size": 32,321  "trial_name": null,322  "trial_params": null323}324