dzungpham/graphcodebert-code-classification

1{2  "best_global_step": null,3  "best_metric": null,4  "best_model_checkpoint": null,5  "epoch": 0.006399590426212722,6  "eval_steps": 1000,7  "global_step": 50,8  "is_hyper_param_search": false,9  "is_local_process_zero": true,10  "is_world_process_zero": true,11  "log_history": [12    {13      "epoch": 0.0012799180852425445,14      "grad_norm": 395827.90625,15      "learning_rate": 1.1523687580025607e-08,16      "loss": 1.0217,17      "step": 1018    },19    {20      "epoch": 0.002559836170485089,21      "grad_norm": 425641.125,22      "learning_rate": 2.4327784891165172e-08,23      "loss": 0.9895,24      "step": 2025    },26    {27      "epoch": 0.0038397542557276334,28      "grad_norm": 322450.875,29      "learning_rate": 3.713188220230474e-08,30      "loss": 1.0297,31      "step": 3032    },33    {34      "epoch": 0.005119672340970178,35      "grad_norm": 362794.28125,36      "learning_rate": 4.9935979513444295e-08,37      "loss": 1.0126,38      "step": 4039    },40    {41      "epoch": 0.006399590426212722,42      "grad_norm": 402867.875,43      "learning_rate": 6.274007682458386e-08,44      "loss": 1.0,45      "step": 5046    }47  ],48  "logging_steps": 10,49  "max_steps": 50,50  "num_input_tokens_seen": 0,51  "num_train_epochs": 1,52  "save_steps": 50,53  "stateful_callbacks": {54    "EarlyStoppingCallback": {55      "args": {56        "early_stopping_patience": 3,57        "early_stopping_threshold": 0.058      },59      "attributes": {60        "early_stopping_patience_counter": 061      }62    },63    "TrainerControl": {64      "args": {65        "should_epoch_stop": false,66        "should_evaluate": false,67        "should_log": false,68        "should_save": true,69        "should_training_stop": true70      },71      "attributes": {}72    }73  },74  "total_flos": 841955377152000.0,75  "train_batch_size": 64,76  "trial_name": null,77  "trial_params": null78}79