| { | |
| "dataset_reader": { | |
| "type": "seq2rel", | |
| "max_length": 512, | |
| "source_token_indexers": { | |
| "tokens": { | |
| "type": "pretrained_transformer", | |
| "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext", | |
| "tokenizer_kwargs": { | |
| "additional_special_tokens": [ | |
| "@ARG@", | |
| "@TRIGGER@" | |
| ], | |
| "do_lower_case": true | |
| } | |
| } | |
| }, | |
| "source_tokenizer": { | |
| "type": "pretrained_transformer", | |
| "add_special_tokens": true, | |
| "max_length": 512, | |
| "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext", | |
| "tokenizer_kwargs": { | |
| "additional_special_tokens": [ | |
| "@ARG@", | |
| "@TRIGGER@" | |
| ], | |
| "do_lower_case": true | |
| } | |
| }, | |
| "target_namespace": "target_tokens", | |
| "target_tokenizer": { | |
| "type": "pretrained_transformer", | |
| "add_special_tokens": false, | |
| "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext", | |
| "tokenizer_kwargs": { | |
| "additional_special_tokens": [ | |
| "@ARG@", | |
| "@TRIGGER@", | |
| "@OSP@", | |
| "@start@", | |
| "@end@", | |
| ";" | |
| ], | |
| "do_lower_case": true | |
| } | |
| } | |
| }, | |
| "model": { | |
| "type": "copynet_seq2rel", | |
| "attention": { | |
| "type": "multihead_attention", | |
| "dropout": 0.1, | |
| "num_heads": 6 | |
| }, | |
| "beam_search": { | |
| "beam_size": 1, | |
| "final_sequence_scorer": { | |
| "type": "length-normalized-sequence-log-prob", | |
| "length_penalty": 1 | |
| }, | |
| "max_steps": 96 | |
| }, | |
| "dropout": 0.1, | |
| "sequence_based_metrics": [ | |
| { | |
| "average": "micro", | |
| "labels": [ | |
| "OSP" | |
| ], | |
| "remove_duplicate_ents": true, | |
| "type": "f1_seq2rel" | |
| } | |
| ], | |
| "source_embedder": { | |
| "token_embedders": { | |
| "tokens": { | |
| "type": "pretrained_transformer", | |
| "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext", | |
| "reinit_modules": 2, | |
| "tokenizer_kwargs": { | |
| "additional_special_tokens": [ | |
| "@ARG@", | |
| "@TRIGGER@" | |
| ], | |
| "do_lower_case": true | |
| } | |
| } | |
| } | |
| }, | |
| "target_embedding_dim": 256, | |
| "target_tokenizer": { | |
| "type": "pretrained_transformer", | |
| "add_special_tokens": false, | |
| "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext", | |
| "tokenizer_kwargs": { | |
| "additional_special_tokens": [ | |
| "@ARG@", | |
| "@TRIGGER@", | |
| "@OSP@", | |
| "@start@", | |
| "@end@", | |
| ";" | |
| ], | |
| "do_lower_case": true | |
| } | |
| }, | |
| "weight_dropout": 0.5 | |
| }, | |
| "train_data_path": "../granular/train_transform.tsv", | |
| "validation_data_path": "../granular/dev_transform.tsv", | |
| "trainer": { | |
| "callbacks": [ | |
| { | |
| "type": "should_validate_callback", | |
| "validation_interval": 1, | |
| "validation_start": 15 | |
| } | |
| ], | |
| "checkpointer": { | |
| "keep_most_recent_by_count": 1 | |
| }, | |
| "grad_norm": 1, | |
| "learning_rate_scheduler": { | |
| "type": "linear_with_warmup", | |
| "warmup_steps": 2906 | |
| }, | |
| "num_epochs": 25, | |
| "num_gradient_accumulation_steps": 1, | |
| "optimizer": { | |
| "type": "huggingface_adamw", | |
| "correct_bias": true, | |
| "eps": 1e-08, | |
| "lr": 0.0004, | |
| "parameter_groups": [ | |
| [ | |
| [ | |
| "transformer_model(?!.*(?:bias|LayerNorm|layer_norm))" | |
| ], | |
| { | |
| "lr": 2e-05, | |
| "weight_decay": 0.01 | |
| } | |
| ], | |
| [ | |
| [ | |
| "transformer_model(?=.*(?:bias|LayerNorm|layer_norm))" | |
| ], | |
| { | |
| "lr": 2e-05, | |
| "weight_decay": 0 | |
| } | |
| ] | |
| ], | |
| "weight_decay": 0 | |
| }, | |
| "use_amp": true, | |
| "validation_metric": "+fscore" | |
| }, | |
| "vocabulary": { | |
| "max_vocab_size": { | |
| "target_tokens": 1 | |
| }, | |
| "tokens_to_add": { | |
| "target_tokens": [ | |
| "@ARG@", | |
| "@TRIGGER@", | |
| "@OSP@", | |
| "@start@", | |
| "@end@", | |
| ";" | |
| ] | |
| } | |
| }, | |
| "data_loader": { | |
| "batch_sampler": { | |
| "type": "bucket", | |
| "batch_size": 4, | |
| "sorting_keys": [ | |
| "source_tokens" | |
| ] | |
| } | |
| }, | |
| "validation_data_loader": { | |
| "batch_sampler": { | |
| "type": "bucket", | |
| "batch_size": 128, | |
| "padding_noise": 0, | |
| "sorting_keys": [ | |
| "source_tokens" | |
| ] | |
| } | |
| } | |
| } |