beratdogan's picture
Upload folder using huggingface_hub
e9b625b verified
Raw
History Blame Contribute Delete
4.09 kB
{
"experiment": "parse",
"name": "parse_xlmr_tr_ota_cv5_fold0",
"n_gpu": 1,
"model": {
"type": "MultiParser",
"args": {
"embeddings_processor": {
"type": "XLMRobertaWrapper",
"args": {
"model_path": "data/pretrained_embeddings/xlm-roberta-base",
"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": 0
},
"paths": {
"train": "/content/steps-otaboun/cv_folds/tr_ota/cv5/fold0_train.conllu",
"dev": "/content/steps-otaboun/cv_folds/tr_ota/cv5/fold0_dev.conllu",
"test_ota": "data/corpora/ota_boun/ota_boun-test-2026.conllu",
"test_tr": "data/corpora/ota_boun/tr_boun-ud-test.conllu"
}
},
"trainer": {
"min_epochs": 15,
"max_epochs": 300,
"early_stop": 15,
"save_dir": "/content/drive/MyDrive/hf_models/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
}
}
}
}