{ "num_channels": 6, "block_out_channels": [ 256, 512, 1024 ], "down_blocks": [ "DownBlock2D", "DownBlock2D", "CrossAttnDownBlock2D" ], "up_blocks": [ "CrossAttnUpBlock2D", "UpBlock2D", "UpBlock2D" ], "layers_per_block": 2, "transformer_layers_per_block": 1, "suggested_image_size": 512 }