{
  "_class_name": "ConsistencyDecoderVAE",
  "_diffusers_version": "0.23.0.dev0",
  "_name_or_path": "openai/consistency-decoder",
  "block_out_channels": [
    128,
    256,
    512,
    512
  ],
  "decoder_add_attention": false,
  "decoder_args": {
    "_use_default_values": [
      "downsample_type",
      "flip_sin_to_cos",
      "mid_block_scale_factor",
      "num_class_embeds",
      "attention_head_dim",
      "sample_size",
      "act_fn",
      "freq_shift",
      "class_embed_type",
      "attn_norm_num_groups",
      "downsample_padding",
      "upsample_type",
      "dropout",
      "center_input_sample"
    ],
    "act_fn": "silu",
    "add_attention": false,
    "attention_head_dim": 8,
    "attn_norm_num_groups": null,
    "block_out_channels": [
      320,
      640,
      1024,
      1024
    ],
    "center_input_sample": false,
    "class_embed_type": null,
    "down_block_types": [
      "ResnetDownsampleBlock2D",
      "ResnetDownsampleBlock2D",
      "ResnetDownsampleBlock2D",
      "ResnetDownsampleBlock2D"
    ],
    "downsample_padding": 1,
    "downsample_type": "conv",
    "dropout": 0.0,
    "flip_sin_to_cos": true,
    "freq_shift": 0,
    "in_channels": 7,
    "layers_per_block": 3,
    "mid_block_scale_factor": 1,
    "norm_eps": 1e-05,
    "norm_num_groups": 32,
    "num_class_embeds": null,
    "num_train_timesteps": 1024,
    "out_channels": 6,
    "resnet_time_scale_shift": "scale_shift",
    "sample_size": null,
    "time_embedding_type": "learned",
    "up_block_types": [
      "ResnetUpsampleBlock2D",
      "ResnetUpsampleBlock2D",
      "ResnetUpsampleBlock2D",
      "ResnetUpsampleBlock2D"
    ],
    "upsample_type": "conv"
  },
  "decoder_block_out_channels": [
    320,
    640,
    1024,
    1024
  ],
  "decoder_down_block_types": [
    "ResnetDownsampleBlock2D",
    "ResnetDownsampleBlock2D",
    "ResnetDownsampleBlock2D",
    "ResnetDownsampleBlock2D"
  ],
  "decoder_downsample_padding": 1,
  "decoder_in_channels": 7,
  "decoder_layers_per_block": 3,
  "decoder_norm_eps": 1e-05,
  "decoder_norm_num_groups": 32,
  "decoder_num_train_timesteps": 1024,
  "decoder_out_channels": 6,
  "decoder_resnet_time_scale_shift": "scale_shift",
  "decoder_time_embedding_type": "learned",
  "decoder_up_block_types": [
    "ResnetUpsampleBlock2D",
    "ResnetUpsampleBlock2D",
    "ResnetUpsampleBlock2D",
    "ResnetUpsampleBlock2D"
  ],
  "encoder_act_fn": "silu",
  "encoder_args": {
    "act_fn": "silu",
    "block_out_channels": [
      128,
      256,
      512,
      512
    ],
    "double_z": true,
    "down_block_types": [
      "DownEncoderBlock2D",
      "DownEncoderBlock2D",
      "DownEncoderBlock2D",
      "DownEncoderBlock2D"
    ],
    "in_channels": 3,
    "layers_per_block": 2,
    "norm_num_groups": 32,
    "out_channels": 4
  },
  "encoder_block_out_channels": [
    128,
    256,
    512,
    512
  ],
  "encoder_double_z": true,
  "encoder_down_block_types": [
    "DownEncoderBlock2D",
    "DownEncoderBlock2D",
    "DownEncoderBlock2D",
    "DownEncoderBlock2D"
  ],
  "encoder_in_channels": 3,
  "encoder_layers_per_block": 2,
  "encoder_norm_num_groups": 32,
  "encoder_out_channels": 4,
  "latent_channels": 4,
  "scaling_factor": 0.18215
}