dzungpham/graphcodebert-code-classification

1{2  "best_global_step": 100,3  "best_metric": 0.5260673147136775,4  "best_model_checkpoint": "output_checkpoints/graphcodebert-vanilla/checkpoint-100",5  "epoch": 0.09784735812133072,6  "eval_steps": 50,7  "global_step": 100,8  "is_hyper_param_search": false,9  "is_local_process_zero": true,10  "is_world_process_zero": true,11  "log_history": [12    {13      "epoch": 0.004892367906066536,14      "grad_norm": 33435.953125,15      "learning_rate": 1.3071895424836603e-07,16      "loss": 0.7006,17      "step": 518    },19    {20      "epoch": 0.009784735812133072,21      "grad_norm": 34831.7421875,22      "learning_rate": 2.9411764705882356e-07,23      "loss": 0.7023,24      "step": 1025    },26    {27      "epoch": 0.014677103718199608,28      "grad_norm": 32986.62109375,29      "learning_rate": 4.5751633986928105e-07,30      "loss": 0.7019,31      "step": 1532    },33    {34      "epoch": 0.019569471624266144,35      "grad_norm": 34257.4921875,36      "learning_rate": 6.209150326797386e-07,37      "loss": 0.7025,38      "step": 2039    },40    {41      "epoch": 0.02446183953033268,42      "grad_norm": 35511.3359375,43      "learning_rate": 7.843137254901962e-07,44      "loss": 0.7032,45      "step": 2546    },47    {48      "epoch": 0.029354207436399216,49      "grad_norm": 33043.3515625,50      "learning_rate": 9.477124183006536e-07,51      "loss": 0.6968,52      "step": 3053    },54    {55      "epoch": 0.03424657534246575,56      "grad_norm": 38538.30078125,57      "learning_rate": 1.111111111111111e-06,58      "loss": 0.6992,59      "step": 3560    },61    {62      "epoch": 0.03913894324853229,63      "grad_norm": 43255.37890625,64      "learning_rate": 1.2745098039215686e-06,65      "loss": 0.6991,66      "step": 4067    },68    {69      "epoch": 0.04403131115459882,70      "grad_norm": 34170.8046875,71      "learning_rate": 1.4379084967320261e-06,72      "loss": 0.6996,73      "step": 4574    },75    {76      "epoch": 0.04892367906066536,77      "grad_norm": 35399.70703125,78      "learning_rate": 1.6013071895424837e-06,79      "loss": 0.6937,80      "step": 5081    },82    {83      "epoch": 0.04892367906066536,84      "eval_accuracy": 0.771,85      "eval_loss": 0.6569265127182007,86      "eval_macro_f1": 0.5082027063884254,87      "eval_precision": 0.6109971743687904,88      "eval_recall": 0.5281120325963375,89      "eval_runtime": 17.5156,90      "eval_samples_per_second": 57.092,91      "eval_steps_per_second": 0.057,92      "step": 5093    },94    {95      "epoch": 0.053816046966731895,96      "grad_norm": 34677.1796875,97      "learning_rate": 1.7647058823529414e-06,98      "loss": 0.6963,99      "step": 55100    },101    {102      "epoch": 0.05870841487279843,103      "grad_norm": 31789.291015625,104      "learning_rate": 1.928104575163399e-06,105      "loss": 0.6971,106      "step": 60107    },108    {109      "epoch": 0.06360078277886497,110      "grad_norm": 30762.345703125,111      "learning_rate": 2.0915032679738565e-06,112      "loss": 0.6946,113      "step": 65114    },115    {116      "epoch": 0.0684931506849315,117      "grad_norm": 29590.083984375,118      "learning_rate": 2.254901960784314e-06,119      "loss": 0.693,120      "step": 70121    },122    {123      "epoch": 0.07338551859099804,124      "grad_norm": 29346.84765625,125      "learning_rate": 2.4183006535947716e-06,126      "loss": 0.6905,127      "step": 75128    },129    {130      "epoch": 0.07827788649706457,131      "grad_norm": 37289.77734375,132      "learning_rate": 2.581699346405229e-06,133      "loss": 0.6942,134      "step": 80135    },136    {137      "epoch": 0.08317025440313111,138      "grad_norm": 33970.94140625,139      "learning_rate": 2.7450980392156867e-06,140      "loss": 0.69,141      "step": 85142    },143    {144      "epoch": 0.08806262230919765,145      "grad_norm": 28671.583984375,146      "learning_rate": 2.9084967320261443e-06,147      "loss": 0.6889,148      "step": 90149    },150    {151      "epoch": 0.09295499021526418,152      "grad_norm": 30156.005859375,153      "learning_rate": 3.071895424836602e-06,154      "loss": 0.6866,155      "step": 95156    },157    {158      "epoch": 0.09784735812133072,159      "grad_norm": 33484.71484375,160      "learning_rate": 3.2352941176470594e-06,161      "loss": 0.6888,162      "step": 100163    },164    {165      "epoch": 0.09784735812133072,166      "eval_accuracy": 0.767,167      "eval_loss": 0.6559821367263794,168      "eval_macro_f1": 0.5260673147136775,169      "eval_precision": 0.6062366452991452,170      "eval_recall": 0.5367285927824044,171      "eval_runtime": 17.3972,172      "eval_samples_per_second": 57.481,173      "eval_steps_per_second": 0.057,174      "step": 100175    }176  ],177  "logging_steps": 5,178  "max_steps": 3066,179  "num_input_tokens_seen": 0,180  "num_train_epochs": 3,181  "save_steps": 100,182  "stateful_callbacks": {183    "EarlyStoppingCallback": {184      "args": {185        "early_stopping_patience": 3,186        "early_stopping_threshold": 0.0187      },188      "attributes": {189        "early_stopping_patience_counter": 0190      }191    },192    "TrainerControl": {193      "args": {194        "should_epoch_stop": false,195        "should_evaluate": false,196        "should_log": false,197        "should_save": true,198        "should_training_stop": false199      },200      "attributes": {}201    }202  },203  "total_flos": 1.3471286034432e+16,204  "train_batch_size": 512,205  "trial_name": null,206  "trial_params": null207}208