config.json

{
  "_class_name": "AutoencoderDC",
  "_diffusers_version": "0.32.0.dev0",
  "attention_head_dim": 32,
  "decoder_act_fns": [
    "relu",
    "relu",
    "relu",
    "silu",
    "silu",
    "silu",
    "silu",
    "silu"
  ],
  "decoder_block_out_channels": [
    128,
    256,
    512,
    512,
    1024,
    1024,
    2048,
    2048
  ],
  "decoder_block_types": [
    "ResBlock",
    "ResBlock",
    "ResBlock",
    "EfficientViTBlock",
    "EfficientViTBlock",
    "EfficientViTBlock",
    "EfficientViTBlock",
    "EfficientViTBlock"
  ],
  "decoder_layers_per_block": [
    0,
    5,
    10,
    2,
    2,
    2,
    2,
    2
  ],
  "decoder_norm_types": [
    "batch_norm",
    "batch_norm",
    "batch_norm",
    "rms_norm",
    "rms_norm",
    "rms_norm",
    "rms_norm",
    "rms_norm"
  ],
  "decoder_qkv_multiscales": [
    [],
    [],
    [],
    [],
    [],
    [],
    [],
    []
  ],
  "downsample_block_type": "pixel_unshuffle",
  "encoder_block_out_channels": [
    128,
    256,
    512,
    512,
    1024,
    1024,
    2048,
    2048
  ],
  "encoder_block_types": [
    "ResBlock",
    "ResBlock",
    "ResBlock",
    "EfficientViTBlock",
    "EfficientViTBlock",
    "EfficientViTBlock",
    "EfficientViTBlock",
    "EfficientViTBlock"
  ],
  "encoder_layers_per_block": [
    0,
    4,
    8,
    2,
    2,
    2,
    2,
    2
  ],
  "encoder_qkv_multiscales": [
    [],
    [],
    [],
    [],
    [],
    [],
    [],
    []
  ],
  "in_channels": 3,
  "latent_channels": 512,
  "scaling_factor": 0.4883,
  "upsample_block_type": "pixel_shuffle"
}