CoolFace
Modelpublic

Portx/v2-extractor-full

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face
0likes36downloads
config.json111 linesDownload Raw Back to root
1{2    "architectures": [3        "MllamaForConditionalGeneration"4    ],5    "image_token_index": 128256,6    "model_type": "mllama",7    "pad_token_id": 128004,8    "text_config": {9        "_attn_implementation_autoset": false,10        "cross_attention_layers": [11            3,12            8,13            13,14            18,15            23,16            28,17            33,18            3819        ],20        "dropout": 0,21        "eos_token_id": [22            128001,23            128008,24            12800925        ],26        "hidden_act": "silu",27        "hidden_size": 4096,28        "initializer_range": 0.02,29        "intermediate_size": 14336,30        "max_position_embeddings": 131072,31        "model_type": "mllama_text_model",32        "num_attention_heads": 32,33        "num_hidden_layers": 40,34        "num_key_value_heads": 8,35        "rms_norm_eps": 1e-05,36        "rope_scaling": {37            "factor": 8.0,38            "high_freq_factor": 4.0,39            "low_freq_factor": 1.0,40            "original_max_position_embeddings": 8192,41            "rope_type": "llama3"42        },43        "rope_theta": 500000.0,44        "torch_dtype": "bfloat16",45        "use_cache": true,46        "vocab_size": 12825647    },48    "torch_dtype": "bfloat16",49    "transformers_version": "4.55.4",50    "unsloth_fixed": true,51    "unsloth_version": "2025.9.11",52    "vision_config": {53        "_attn_implementation_autoset": false,54        "attention_heads": 16,55        "hidden_act": "gelu",56        "hidden_size": 1280,57        "image_size": 560,58        "initializer_range": 0.02,59        "intermediate_layers_indices": [60            3,61            7,62            15,63            23,64            3065        ],66        "intermediate_size": 5120,67        "max_num_tiles": 4,68        "model_type": "mllama_vision_model",69        "norm_eps": 1e-05,70        "num_channels": 3,71        "num_global_layers": 8,72        "num_hidden_layers": 32,73        "patch_size": 14,74        "supported_aspect_ratios": [75            [76                1,77                178            ],79            [80                1,81                282            ],83            [84                1,85                386            ],87            [88                1,89                490            ],91            [92                2,93                194            ],95            [96                2,97                298            ],99            [100                3,101                1102            ],103            [104                4,105                1106            ]107        ],108        "torch_dtype": "bfloat16",109        "vision_output_dim": 7680110    }111}