| { | |
| "architectures": ["LaTeXOCRModel"], | |
| "model_type": "latex_ocr", | |
| "processor_class": "LaTeXOCRProcessor", | |
| "auto_map": { | |
| "AutoConfig": "configuration_latex_ocr.LaTeXOCRConfig", | |
| "AutoProcessor": "processing_latex_ocr.LaTeXOCRProcessor", | |
| "AutoImageProcessor": "image_processing_latex_ocr.LaTeXOCRImageProcessor", | |
| "AutoModel": "modeling_latex_ocr.LaTeXOCRModel", | |
| "AutoTokenizer": ["tokenization_latex_ocr.LaTeXTokenizer", null], | |
| "Pipeline": "pipeline_latex_ocr.LaTeXOCRPipeline" | |
| }, | |
| "decoder_weights_tied": false, | |
| "tie_word_embeddings": false, | |
| "pad_token_id": 0, | |
| "bos_token_id": 2, | |
| "eos_token_id": 3, | |
| "decoder_start_token_id": 2, | |
| "navit_dim": 512, | |
| "vision_hidden_size": 512, | |
| "llm_hidden_size": 512, | |
| "vocab_size": 2046, | |
| "patch_size": 16, | |
| "image_height": 64, | |
| "max_visual_tokens": 256, | |
| "max_new_tokens": 200, | |
| "num_beams": 4, | |
| "decoder_arch": { | |
| "vocab_size": 2046, | |
| "pad_id": 0, | |
| "bos_id": 2, | |
| "eos_id": 3, | |
| "d_model": 512, | |
| "n_heads": 8, | |
| "n_layers": 6, | |
| "d_ff": 1408, | |
| "dropout": 0.1, | |
| "max_seq_len": 200, | |
| "rope_theta": 10000.0, | |
| "tie_weights": false | |
| } | |
| } |