{
  "architectures": [
    "OmnivoreForJointClassification"
  ],
  "attention_dropout_rate": 0.0,
  "depth_mode": "summed_rgb_d_tokens",
  "depths": [
    2,
    2,
    18,
    2
  ],
  "drop_path_rate": 0.3,
  "dropout_rate": 0.0,
  "embed_dim": 96,
  "frozen_stages": -1,
  "head_dim_in": 768,
  "initializer_range": 0.02,
  "input_channels": 3,
  "mlp_ratio": 4.0,
  "model_type": "omnivore",
  "num_heads": [
    3,
    6,
    12,
    24
  ],
  "num_image_labels": 1000,
  "num_rgbd_labels": 19,
  "num_video_labels": 400,
  "patch_norm": true,
  "patch_size": [
    2,
    4,
    4
  ],
  "qk_scale": null,
  "qkv_bias": true,
  "torch_dtype": "float32",
  "transformers_version": "4.21.0.dev0",
  "window_size": [
    8,
    7,
    7
  ]
}