{
  "architectures": [
    "2D"
  ],
  "attention_head_dim": 8,
  "block_out_channels": [
    320,
    640,
    1280,
    1280
  ],
  "cross_attention_dim": 768,
  "down_block_types": [
    "CrossAttnDownBlock2D",
    "CrossAttnDownBlock2D",
    "CrossAttnDownBlock2D",
    "DownBlock2D"
  ],
  "dropout": 0.1,
  "in_channels": 4,
  "layers_per_block": 2,
  "out_channels": 4,
  "sample_size": 64,
  "transformers_version": "4.21.0.dev0",
  "up_block_types": [
    "UpBlock2D",
    "CrossAttnUpBlock2D",
    "CrossAttnUpBlock2D",
    "CrossAttnUpBlock2D"
  ]
}