{
  "_class_name": "ControlNetModel",
  "_diffusers_version": "0.26.1",
  "act_fn": "silu",
  "attention_head_dim": 64,
  "block_out_channels": [
    320,
    640,
    1280,
    1280
  ],
  "class_embed_type": null,
  "conditioning_embedding_out_channels": [
    32,
    96,
    256,
    512
  ],
  "controlnet_conditioning_channel_order": "rgb",
  "cross_attention_dim": 1024,
  "down_block_types": [
    "CrossAttnDownBlock3D",
    "CrossAttnDownBlock3D",
    "CrossAttnDownBlock3D",
    "DownBlock3D"
  ],
  "downsample_controlnet_cond": true,
  "downsample_padding": 1,
  "flip_sin_to_cos": true,
  "frame_expansion": "none",
  "freq_shift": 0,
  "global_pool_conditions": false,
  "in_channels": 4,
  "layers_per_block": 2,
  "merging_mode": "addition",
  "mid_block_scale_factor": 1,
  "norm_eps": 1e-05,
  "norm_num_groups": 32,
  "num_class_embeds": null,
  "num_frames": 8,
  "num_frames_conditioning": 8,
  "num_tranformers": 1,
  "only_cross_attention": false,
  "projection_class_embeddings_input_dim": null,
  "resnet_time_scale_shift": "default",
  "upcast_attention": false,
  "use_controlnet_mask": false,
  "use_image_embedding": false,
  "use_image_encoder_normalization": false,
  "use_image_tokens": false,
  "use_linear_projection": false,
  "use_repeat_context_img": true,
  "zero_conv_mode": "Identity"
}