{
  "_class_name": "UNet3DConditionModel",
  "_diffusers_version": "0.27.2",
  "act_fn": "silu",
  "attention_head_dim": 64,
  "block_out_channels": [
    64,
    128,
    128,
    256
  ],
  "cross_attention_dim": 8,
  "down_block_types": [
    "CrossAttnDownBlock3D",
    "CrossAttnDownBlock3D",
    "CrossAttnDownBlock3D",
    "DownBlock3D"
  ],
  "downsample_padding": 1,
  "in_channels": 1,
  "layers_per_block": 2,
  "mid_block_scale_factor": 1,
  "norm_eps": 1e-05,
  "norm_num_groups": 32,
  "num_attention_heads": null,
  "out_channels": 1,
  "sample_size": [
    32,
    32,
    32
  ],
  "up_block_types": [
    "UpBlock3D",
    "CrossAttnUpBlock3D",
    "CrossAttnUpBlock3D",
    "CrossAttnUpBlock3D"
  ]
}