vae/config.json

{
    "_class_name": "AutoencoderKL",
    "block_out_channels": [
        128,
        256,
        512,
        512
    ],
    "in_channels": 3,
    "latent_channels": 4,
    "layers_per_block": 2,
    "norm_num_groups": 32,
    "out_channels": 3,
    "sample_size": 512
}