{"model_type": "multimodal-bias-classifier", "text_encoder_id_or_path": "bert-base-uncased", "image_encoder_id_or_path": "resnet34", "projection_dim": 768, "fusion_method": "concat", "num_classes": 1, "proj_dropout": 0.1, "fusion_dropout": 0.1, "hidden_size": 768, "save_components": ["resnet_encoder", "text_encoder", "fusion_layer", "classifier"], "exclude_components": ["clip_text_encoder", "clip_image_encoder"]}