| { | |
| "model_select": "distilgpt2", | |
| "dataset_name": "source_code", | |
| "per_gpu_train_batch_size": 4, | |
| "dev_batch_size": 8, | |
| "num_epochs_train": 5, | |
| "max_seq_length": 256, | |
| "lr": 2e-05, | |
| "warmup_ratio": 0.2, | |
| "early_stop": 3, | |
| "scheduler": "warmuplinear", | |
| "seed": 122, | |
| "accumulation_steps": 1, | |
| "n_gpu": 4, | |
| "visiable_device": "0", | |
| "evaluation_steps": 200, | |
| "wandb_project_name": "code_generate", | |
| "restore_training": false, | |
| "with_wandb": true, | |
| "wandb_run_name": "model/distilgpt2_fine_tuned_coder" | |
| } |