controlnet/config.json

{
  "_class_name": "BEVControlNetModel",
  "_diffusers_version": "0.17.1",
  "act_fn": "silu",
  "attention_head_dim": 8,
  "bbox_embedder_cls": "dreamforge.networks.bbox_embedder.ContinuousBBoxWithTextEmbedding",
  "bbox_embedder_param": {
    "class_token_dim": 768,
    "embedder_num_freq": 4,
    "minmax_normalize": false,
    "mode": "all-xyz",
    "n_classes": 10,
    "proj_dims": [
      768,
      512,
      512,
      768
    ],
    "trainable_class_token": false,
    "use_text_encoder_init": true
  },
  "block_out_channels": [
    320,
    640,
    1280,
    1280
  ],
  "cam_embedder_param": {
    "include_input": true,
    "input_dims": 3,
    "log_sampling": true,
    "num_freqs": 4
  },
  "camera_in_dim": 189,
  "camera_out_dim": 768,
  "canvas_conditioning_channels": 14,
  "class_embed_type": null,
  "conditioning_embedding_out_channels": [
    16,
    32,
    96,
    256
  ],
  "controlnet_conditioning_channel_order": "rgb",
  "cross_attention_dim": 768,
  "down_block_types": [
    "CrossAttnDownBlock2D",
    "CrossAttnDownBlock2D",
    "CrossAttnDownBlock2D",
    "DownBlock2D"
  ],
  "downsample_padding": 1,
  "drop_cam_num": 6,
  "drop_cam_with_box": false,
  "drop_cond_ratio": 0.25,
  "flip_sin_to_cos": true,
  "freq_shift": 0,
  "global_pool_conditions": false,
  "in_channels": 4,
  "layers_per_block": 2,
  "map_embedder_cls": null,
  "map_embedder_param": null,
  "map_size": [
    4,
    200,
    200
  ],
  "mid_block_scale_factor": 1,
  "norm_eps": 1e-05,
  "norm_num_groups": 32,
  "num_class_embeds": null,
  "only_cross_attention": false,
  "projection_class_embeddings_input_dim": null,
  "resnet_time_scale_shift": "default",
  "uncond_cam_in_dim": [
    3,
    7
  ],
  "upcast_attention": false,
  "use_linear_projection": false,
  "use_uncond_map": null,
  "with_layout_canvas": true
}