{
  "_class_name": "ControlNetModel",
  "_diffusers_version": "0.17.0.dev0",
  "_name_or_path": "gdrive/MIDI-AudioLDM/AudioLDM-m-text-ft-urmp-maestro-slakh-audio-encoder-no-drums-no-mix/checkpoint-10000",
  "act_fn": "silu",
  "attention_head_dim": 8,
  "block_out_channels": [
    192,
    384,
    576,
    960
  ],
  "class_embed_type": "simple_projection",
  "class_embeddings_concat": true,
  "conditioning_channels": 1,
  "conditioning_embedding_out_channels": [
    16,
    32,
    96,
    256
  ],
  "controlnet_conditioning_channel_order": "rgb",
  "cross_attention_dim": [
    192,
    384,
    576,
    960
  ],
  "down_block_types": [
    "DownBlock2D",
    "CrossAttnDownBlock2D",
    "CrossAttnDownBlock2D",
    "CrossAttnDownBlock2D"
  ],
  "downsample_padding": 1,
  "flip_sin_to_cos": true,
  "freq_shift": 0,
  "global_pool_conditions": false,
  "in_channels": 8,
  "layers_per_block": 2,
  "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": 512,
  "resnet_time_scale_shift": "default",
  "upcast_attention": false,
  "use_linear_projection": false
}