| { | |
| "pretrained_model_name_or_path": "", | |
| "path_relm": "relm-m0.3.bin", | |
| "path_train": "csc_public.train.json", | |
| "path_dev": "csc_public.dev.json", | |
| "path_tet": "csc_public.tet.json", | |
| "model_save_path": "../output/relm_v1", | |
| "task_name": "relm_csc", | |
| "do_lower_case": true, | |
| "do_train": true, | |
| "do_eval": true, | |
| "do_test": true, | |
| "gradient_accumulation_steps": 4, | |
| "warmup_proportion": 0.1, | |
| "num_warmup_steps": null, | |
| "max_train_steps": null, | |
| "num_train_epochs": 3, | |
| "train_batch_size": 8, | |
| "eval_batch_size": 8, | |
| "learning_rate": 3e-05, | |
| "max_seq_length": 256, | |
| "max_grad_norm": 1.0, | |
| "weight_decay": 0.0005, | |
| "save_steps": 1000, | |
| "anchor": null, | |
| "seed": 42, | |
| "lr_scheduler_type": "cosine", | |
| "loss_type": "focal_loss", | |
| "mask_mode": "noerror", | |
| "loss_det_rate": 0.3, | |
| "prompt_length": 0, | |
| "mask_rate": 0.3, | |
| "threshold": 0.5, | |
| "flag_dynamic_encode": false, | |
| "flag_loss_period": false, | |
| "flag_cpo_loss": false, | |
| "flag_fast_tokenizer": true, | |
| "flag_pin_memory": true, | |
| "flag_train": false, | |
| "flag_fp16": false, | |
| "flag_cuda": true, | |
| "flag_skip": true, | |
| "flag_mft": true, | |
| "num_workers": 0, | |
| "CUDA_VISIBLE_DEVICES": "0", | |
| "USE_TORCH": "1" | |
| } |