{
  "in_channels": 3,
  "model_channels": 256,
  "out_channels": 6,
  "num_res_blocks": 3,
  "attention_resolutions": [
    2,
    4,
    8
  ],
  "channel_mult": [
    1,
    2,
    2,
    4
  ],
  "time_embed_dim": 512,
  "use_scale_shift_norm": true,
  "conv_resample": false
}