{
  "attn_resolutions": [
    16
  ],
  "hidden_channels": 128,
  "channel_mult": [
    1,
    1,
    2,
    2,
    4
  ],
  "dropout": 0.0,
  "quantized_embed_dim": 256,
  "num_channels": 3,
  "num_embeddings": 1024,
  "num_res_blocks": 2,
  "resample_with_conv": true,
  "resolution": 256,
  "z_channels": 256
}