| { | |
| "embed_dim": 768, | |
| "vision_cfg": { | |
| "image_size": 224, | |
| "layers": 12, | |
| "width": 768, | |
| "patch_size": 16 | |
| }, | |
| "text_cfg": { | |
| "context_length": 280, | |
| "vocab_size": 49411, | |
| "width": 768, | |
| "heads": 12, | |
| "layers": 12, | |
| "tokenizer_kwargs": { | |
| "additional_special_tokens": [ | |
| "<img_placehold>", | |
| "<start_of_img>", | |
| "<end_of_img>" | |
| ] | |
| } | |
| }, | |
| "pool_project_cfg": { | |
| "pool_proj_type": "attn", | |
| "input_dim": 768, | |
| "output_dim": 768, | |
| "attn_num_heads": 12 | |
| }, | |
| "special_token_ids": { | |
| "<start_of_text>": 49406, | |
| "<end_of_text>": 49407, | |
| "<img_placehold>": 49408, | |
| "<start_of_img>": 49409, | |
| "<end_of_img>": 49410 | |
| }, | |
| "custom_text": true | |
| } |