config.json

{
  "adapter_config": {
    "attention_dropout": 0.0,
    "float32_attention": true,
    "head_dim": 72,
    "hidden_act": "silu",
    "hidden_size": 1152,
    "image_feature_dropout": 0.0,
    "image_padding_embed": null,
    "initializer_range": 0.02,
    "intermediate_size": 18944,
    "model_type": "",
    "num_attention_heads": 16,
    "num_key_value_heads": 16,
    "residual_dropout": 0.0,
    "text_hidden_size": 3584,
    "vit_layers": [
      -3,
      -9
    ]
  },
  "architectures": [
    "MolmoActForActionReasoning"
  ],
  "auto_map": {
    "AutoConfig": "configuration_molmoact.MolmoActConfig",
    "AutoModelForImageTextToText": "modeling_molmoact.MolmoActForActionReasoning"
  },
  "image_patch_id": 152066,
  "initializer_range": 0.02,
  "llm_config": {
    "additional_vocab_size": 128,
    "attention_dropout": 0.0,
    "embedding_dropout": 0.0,
    "head_dim": 128,
    "hidden_act": "silu",
    "hidden_size": 3584,
    "initializer_range": 0.02,
    "intermediate_size": 18944,
    "layer_norm_eps": 1e-06,
    "max_position_embeddings": 4096,
    "model_type": "molmoact_llm",
    "norm_after": false,
    "num_attention_heads": 28,
    "num_hidden_layers": 28,
    "num_key_value_heads": 4,
    "qk_norm_type": "olmo",
    "qkv_bias": true,
    "residual_dropout": 0.0,
    "rope_scaling": null,
    "rope_theta": 1000000.0,
    "use_cache": true,
    "use_qk_norm": false,
    "vocab_size": 152064
  },
  "model_type": "molmoact",
  "n_action_bins": 256,
  "norm_stats": {
    "molmoact": {
      "action": {
        "max": [
          0.06042003631591797,
          0.09417290985584259,
          0.07019275426864624,
          0.2616892158985138,
          0.11751057207584381,
          0.16968433558940887,
          1.0
        ],
        "mean": [
          0.0005706787342205644,
          0.0002448957529850304,
          -3.5987635783385485e-05,
          0.00021597897284664214,
          -0.0004896928439848125,
          -0.000241481073317118,
          0.5570635199546814
        ],
        "min": [
          -0.07434078305959702,
          -0.07339745759963989,
          -0.06539416313171387,
          -0.1688285619020462,
          -0.10289879888296127,
          -0.2667275667190552,
          0.0
        ],
        "q01": [
          -0.01538565568625927,
          -0.021047022193670273,
          -0.01688069850206375,
          -0.044314172118902206,
          -0.03890235349535942,
          -0.04788423702120781,
          0.0
        ],
        "q99": [
          0.014661382883787155,
          0.026515591889619827,
          0.021398313343524933,
          0.04216696694493294,
          0.03401297703385353,
          0.04957397282123566,
          1.0
        ],
        "std": [
          0.005207270849496126,
          0.007506529800593853,
          0.006415561307221651,
          0.013248044066131115,
          0.010928540490567684,
          0.014873150736093521,
          0.49715080857276917
        ]
      },
      "num_entries": 1560068
    }
  },
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.52.3",
  "use_cache": true,
  "vit_config": {
    "attention_dropout": 0.0,
    "float32_attention": true,
    "head_dim": 72,
    "hidden_act": "gelu_pytorch_tanh",
    "hidden_size": 1152,
    "image_default_input_size": [
      378,
      378
    ],
    "image_num_pos": 729,
    "image_patch_size": 14,
    "initializer_range": 0.02,
    "intermediate_size": 4304,
    "layer_norm_eps": 1e-06,
    "model_type": "molmoact_vit",
    "num_attention_heads": 16,
    "num_hidden_layers": 27,
    "num_key_value_heads": 16,
    "patch_bias": true,
    "pre_layernorm": false,
    "residual_dropout": 0.0,
    "use_cls_token": false
  }
}