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": {
    "libero_spatial_no_noops_modified": {
      "action": {
        "max": [
          0.9375,
          0.9375,
          0.9375,
          0.1971428543329239,
          0.33642858266830444,
          0.375,
          1.0
        ],
        "mean": [
          0.15312479436397552,
          0.13707277178764343,
          -0.15526802837848663,
          -0.005176450591534376,
          -0.01120874285697937,
          -0.020194264128804207,
          0.4578818082809448
        ],
        "min": [
          -0.9375,
          -0.9375,
          -0.9375,
          -0.1875,
          -0.3675000071525574,
          -0.36000001430511475,
          0.0
        ],
        "q01": [
          -0.7454732114076613,
          -0.6616071462631226,
          -0.9375,
          -0.1071428582072258,
          -0.20678570866584778,
          -0.1842857152223587,
          0.0
        ],
        "q99": [
          0.9375,
          0.8758928775787354,
          0.9321428537368774,
          0.1039285734295845,
          0.17678570747375488,
          0.14571428298950195,
          1.0
        ],
        "std": [
          0.41272708773612976,
          0.34724321961402893,
          0.50869220495224,
          0.037266165018081665,
          0.07244449853897095,
          0.05762382969260216,
          0.49827873706817627
        ]
      },
      "num_trajectories": 432,
      "num_transitions": 52970,
      "proprio": {
        "max": [
          0.1759040206670761,
          0.3904820382595062,
          1.3290715217590332,
          3.4566118717193604,
          1.2268599271774292,
          1.0429412126541138,
          0.0,
          0.041053611785173416,
          0.000775813648942858
        ],
        "mean": [
          -0.024462558329105377,
          0.106529600918293,
          1.0580483675003052,
          3.0628468990325928,
          -0.10464039444923401,
          0.08307311683893204,
          0.0,
          0.01995457336306572,
          -0.020162804052233696
        ],
        "min": [
          -0.3095473051071167,
          -0.29250794649124146,
          0.9095591306686401,
          2.497488260269165,
          -1.8006486892700195,
          -0.7207611203193665,
          0.0,
          -0.0004703797458205372,
          -0.041536275297403336
        ],
        "q01": [
          -0.2727657300233841,
          -0.23721413239836692,
          0.9160063165426254,
          2.77949666261673,
          -1.3187511622905732,
          -0.41989982962608335,
          0.0,
          0.001503719249740243,
          -0.03989770736545324
        ],
        "q99": [
          0.13529365032911292,
          0.3629165390133857,
          1.2862326657772063,
          3.2829698753356933,
          0.9332760351896285,
          0.6325724506378171,
          0.0,
          0.039933966137468815,
          -0.001671919699292631
        ],
        "std": [
          0.1101478561758995,
          0.13784688711166382,
          0.1044282391667366,
          0.10451053828001022,
          0.4112098217010498,
          0.2176690548658371,
          0.0,
          0.017260896041989326,
          0.0171116404235363
        ]
      }
    }
  },
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "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
  }
}