{ "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext", "model_type": "StableAphasiaClassifier", "num_labels": 9, "architectures": ["StableAphasiaClassifier"], "_name_or_path": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext", "problem_type": "single_label_classification", "aphasia_types_mapping": { "BROCA": 0, "TRANSMOTOR": 1, "NOTAPHASICBYWAB": 2, "CONDUCTION": 3, "WERNICKE": 4, "ANOMIC": 5, "GLOBAL": 6, "ISOLATION": 7, "TRANSSENSORY": 8 }, "id2label": { "0": "BROCA", "1": "TRANSMOTOR", "2": "NOTAPHASICBYWAB", "3": "CONDUCTION", "4": "WERNICKE", "5": "ANOMIC", "6": "GLOBAL", "7": "ISOLATION", "8": "TRANSSENSORY" }, "label2id": { "BROCA": 0, "TRANSMOTOR": 1, "NOTAPHASICBYWAB": 2, "CONDUCTION": 3, "WERNICKE": 4, "ANOMIC": 5, "GLOBAL": 6, "ISOLATION": 7, "TRANSSENSORY": 8 }, "model_config": { "max_length": 512, "hidden_size": 768, "pos_vocab_size": 150, "pos_emb_dim": 64, "grammar_dim": 3, "grammar_hidden_dim": 64, "duration_hidden_dim": 128, "prosody_dim": 32, "num_attention_heads": 8, "attention_dropout": 0.3, "classifier_hidden_dims": [512, 256], "dropout_rate": 0.3, "activation_fn": "tanh" }, "training_config": { "learning_rate": 0.0005, "weight_decay": 0.01, "warmup_ratio": 0.1, "batch_size": 10, "gradient_accumulation_steps": 4, "use_focal_loss": true, "focal_alpha": 1.0, "focal_gamma": 2.0, "use_mixup": false, "mixup_alpha": 0.2, "use_label_smoothing": true, "label_smoothing": 0.1 }, "adaptive_lr_config": { "adaptive_lr": true, "lr_patience": 3, "lr_factor": 0.8, "lr_increase_factor": 1.2, "min_lr": 1e-06, "max_lr": 0.001, "oscillation_amplitude": 0.1 }, "tokenizer_config": { "pad_token": "[PAD]", "unk_token": "[UNK]", "sep_token": "[SEP]", "cls_token": "[CLS]", "mask_token": "[MASK]", "additional_special_tokens": [ "[DIALOGUE]", "[TURN]", "[PAUSE]", "[REPEAT]", "[HESITATION]" ] }, "torch_dtype": "float32", "transformers_version": "4.36.0", "_commit_hash": null, "auto_map": { "AutoModel": "StableAphasiaClassifier" } }