config.json

{
  "architectures": [
    "MllamaForConditionalGeneration"
  ],
  "dtype": "float32",
  "image_token_index": 4,
  "model_type": "mllama",
  "text_config": {
    "bos_token_id": 128000,
    "cross_attention_layers": [
      1
    ],
    "dropout": 0.0,
    "eos_token_id": 128001,
    "hidden_act": "gelu",
    "hidden_size": 32,
    "initializer_range": 0.02,
    "intermediate_size": 37,
    "max_position_embeddings": 512,
    "model_type": "mllama_text_model",
    "num_attention_heads": 4,
    "num_hidden_layers": 2,
    "num_key_value_heads": 4,
    "pad_token_id": 128004,
    "rms_norm_eps": 1e-05,
    "rope_parameters": {
      "rope_theta": 500000.0,
      "rope_type": "default"
    },
    "tie_word_embeddings": false,
    "use_cache": true,
    "vocab_size": 128257
  },
  "transformers_version": "5.16.0.dev0",
  "vision_config": {
    "attention_heads": 4,
    "dropout": 0.1,
    "hidden_act": "gelu",
    "hidden_size": 16,
    "image_size": 30,
    "initializer_range": 0.02,
    "intermediate_layers_indices": [
      0
    ],
    "intermediate_size": 37,
    "max_num_tiles": 4,
    "model_type": "mllama_vision_model",
    "norm_eps": 1e-05,
    "num_channels": 3,
    "num_global_layers": 2,
    "num_hidden_layers": 2,
    "patch_size": 2,
    "projection_dim": 32,
    "supported_aspect_ratios": [
      [
        1,
        1
      ],
      [
        1,
        2
      ],
      [
        1,
        3
      ],
      [
        1,
        4
      ],
      [
        2,
        1
      ],
      [
        2,
        2
      ],
      [
        3,
        1
      ],
      [
        4,
        1
      ]
    ],
    "vision_output_dim": 32
  }
}