| { | |
| "model_wrapper": { | |
| "type": "token_classification", | |
| "num_labels": 3 | |
| }, | |
| "label_mapper": { | |
| "type": "answer-extraction" | |
| }, | |
| "args": { | |
| "type": "default", | |
| "disable_tqdm": true, | |
| "do_eval": true, | |
| "do_train": true, | |
| "eval_steps": 1824, | |
| "evaluation_strategy": "steps", | |
| "fp16": true, | |
| "fp16_backend": "amp", | |
| "gradient_accumulation_steps": 1, | |
| "greater_is_better": true, | |
| "label_names": [ | |
| "labels" | |
| ], | |
| "logging_dir": "/opt/ml/checkpoints/logs", | |
| "logging_first_step": true, | |
| "logging_steps": 1824, | |
| "lr_scheduler_type": "linear", | |
| "metric_for_best_model": "eval_ANS_f1", | |
| "no_cuda": false, | |
| "num_train_epochs": 3, | |
| "output_dir": "/opt/ml/checkpoints", | |
| "per_device_eval_batch_size": 10, | |
| "per_device_train_batch_size": 4, | |
| "result_dir": "/opt/ml/model", | |
| "save_steps": 1824, | |
| "save_total_limit": null, | |
| "warmup_steps": 985 | |
| }, | |
| "compute_metrics": { | |
| "type": "seqeval" | |
| }, | |
| "dataset_loader": { | |
| "dataset_reader": { | |
| "path": "squad_qa_test_fixture" | |
| }, | |
| "data_processor": { | |
| "type": "squad-answer-extraction" | |
| }, | |
| "data_adapter": { | |
| "type": "answer-extraction" | |
| } | |
| }, | |
| "dev_file_path": "/opt/ml/input/data/validation/dev.json", | |
| "optimizer": { | |
| "type": "huggingface_adamw", | |
| "eps": 1e-06, | |
| "lr": 3e-05, | |
| "parameter_groups": [ | |
| [ | |
| [ | |
| "bias", | |
| "LayerNorm\\\\.weight", | |
| "layer_norm\\\\.weight" | |
| ], | |
| { | |
| "weight_decay": 0 | |
| } | |
| ] | |
| ], | |
| "weight_decay": 0.01 | |
| }, | |
| "pretrained_model_name_or_path": "roberta-large", | |
| "tokenizer_wrapper": { | |
| "type": "answer-extraction" | |
| }, | |
| "train_file_path": "/opt/ml/input/data/training/train.json" | |
| } |