{
    "in_channels": 16,
    "out_channels": 3,
    "up_block_types": [
        "DecoderUpBlock2D",
        "DecoderUpBlock2D",
        "DecoderUpBlock2D",
        "DecoderUpBlock2D"
    ],
    "block_out_channels": [
        64,
        128,
        256,
        256
    ],
    "layers_per_block": 2,
    "norm_num_groups": 32,
    "act_fn": "hardswish",
    "mid_block_add_attention": false,
    "conv_in_module": "Conv2d",
    "conv_in_dw_bias": true,
    "conv_in_pw_bias": false,
    "conv_out_module": "Conv2d",
    "conv_out_dw_bias": true,
    "conv_out_pw_bias": false,
    "use_mid_block": false,
    "mid_block_type": "DecoderUNetMidBlock2D",
    "mid_block_use_additional_resnet": true,
    "resnet_middle_expansion": null,
    "resnet_module": "DecoderResnetBlock2D",
    "resnet_conv_module": "DecoderSepConv2d",
    "attn_module": "Attention",
    "attn_processor_type": "AttnProcessor2_0",
    "kv_heads": 1,
    "qk_norm": "layer_norm",
    "layers_per_blocks": [
        3,
        3,
        2,
        1
    ],
    "backward_output_channels": true,
    "upsample_module": "Upsample2D",
    "upsample_conv_module": "Conv2d",
    "resnet_dw_bias": true,
    "resnet_pw_bias": true
}