{ | |
"lr_encoder": "1e-5", | |
"lr_others": "5e-5", | |
"num_steps": 60000, | |
"warmup_ratio": 0.1, | |
"train_batch_size": 4, | |
"gradient_accumulation_steps": 2, | |
"eval_every": 2500, | |
"max_width": 1, | |
"model_name": "microsoft/deberta-v3-large", | |
"fine_tune": true, | |
"subtoken_pooling": "first", | |
"hidden_size": 768, | |
"span_mode": "marker", | |
"dropout": 0.4, | |
"root_dir": "ablation_backbone", | |
"train_data": "NuMinds_custom_data_mix.json", | |
"prev_path": "none", | |
"size_sup": -1, | |
"max_types": 25, | |
"shuffle_types": true, | |
"random_drop": true, | |
"max_neg_type_ratio": 1, | |
"max_len": 384, | |
"name": "large", | |
"log_dir": "logs" | |
} |