| { | |
| "experiment": "ota_boun", | |
| "name": "bert-base-turkish-cased", | |
| "n_gpu": 1, | |
| "model": { | |
| "type": "MultiParser", | |
| "args": { | |
| "embeddings_processor": { | |
| "type": "BertWrapper", | |
| "args": { | |
| "model_path": "data/pretrained_embeddings/bert-base-turkish-cased-nonpretrained", | |
| "fine_tune": true, | |
| "hidden_dropout": 0.2, | |
| "attn_dropout": 0.2, | |
| "output_dropout": 0.5, | |
| "scalar_mix_layer_dropout": 0.1, | |
| "token_mask_prob": 0.15 | |
| } | |
| }, | |
| "outputs": { | |
| "heads": { | |
| "type": "ArcScorer", | |
| "args": { | |
| "scorer_class": "DeepBiaffineScorer", | |
| "head_mode": "single_head", | |
| "hidden_size": 768, | |
| "dropout": 0.33, | |
| "vocab": { | |
| "type": "IntegerVocab" | |
| } | |
| } | |
| }, | |
| "labels": { | |
| "type": "DependencyClassifier", | |
| "args": { | |
| "scorer_class": "DeepBiaffineScorer", | |
| "hidden_size": 256, | |
| "dropout": 0.33, | |
| "vocab": { | |
| "type": "BasicVocab", | |
| "args": { | |
| "vocab_filename": "data/corpora/ota_boun/vocab/basic.vocab" | |
| } | |
| } | |
| } | |
| } | |
| }, | |
| "post_processors": [ | |
| { | |
| "type": "FactorizedMSTPostProcessor", | |
| "args": { | |
| "annotation_ids": [ | |
| "heads", | |
| "labels" | |
| ] | |
| } | |
| } | |
| ] | |
| } | |
| }, | |
| "data_loaders": { | |
| "type": "BucketedCONLLLoader", | |
| "args": { | |
| "annotation_layers": { | |
| "heads": { | |
| "type": "TagSequence", | |
| "source_column": 6, | |
| "args": { | |
| "ignore_root": true | |
| } | |
| }, | |
| "labels": { | |
| "type": "DependencyMatrix", | |
| "source_column": [ | |
| 6, | |
| 7 | |
| ], | |
| "args": { | |
| "ignore_non_relations": true | |
| } | |
| } | |
| }, | |
| "batch_size": 32, | |
| "bucket_size": 256, | |
| "max_tokens_per_batch": 156800, | |
| "num_workers": 2 | |
| }, | |
| "paths": { | |
| "train": "data/corpora/ota_boun/ota_boun-ud-train.conllu", | |
| "dev": "data/corpora/ota_boun/ota_boun-ud-test.conllu", | |
| "test": "data/corpora/ota_boun/tr_boun-ud-test.conllu" | |
| } | |
| }, | |
| "trainer": { | |
| "min_epochs": 15, | |
| "max_epochs": 300, | |
| "early_stop": 15, | |
| "save_dir": "data/saved_models/", | |
| "save_period": 20, | |
| "verbosity": 2, | |
| "validation_criterion": { | |
| "metrics": { | |
| "heads": "fscore", | |
| "labels": "fscore" | |
| }, | |
| "weighting": "multiplicative" | |
| }, | |
| "optimizer": { | |
| "type": "AdamW", | |
| "args": { | |
| "lr": 4e-05, | |
| "weight_decay": 0.0 | |
| } | |
| }, | |
| "lr_scheduler": { | |
| "type": "LambdaLR", | |
| "args": { | |
| "lr_lambda": "SqrtSchedule(400)" | |
| } | |
| }, | |
| "loss": { | |
| "type": "CrossEntropyLoss", | |
| "args": { | |
| "ignore_index": -1 | |
| } | |
| } | |
| } | |
| } |