VTP.Config(
        encoder=Encoder.Config(
            patch_size=16,
            channels_hidden=384,
            num_layers=12,
            heads=6,
            expansion=4.0,
            channels_out=64,
            checkpoint=HubFile.Config(
                repo_id='MiniMaxAI/VTP-Small-f16d64',
                filename='model.safetensors',
                revision='378967941e66f7f09a9e10218aa8710a930cc635',
                sha256='5442083e078aadc65e51c7893d10177e360139866d4e09ab955b1c7b563eeaee'
            )
        ),
        decoder=Decoder.Config(
            channels_in=64,
            channels_hidden=384,
            num_layers=12,
            heads=6,
            expansion=4.0,
            upscale_factor=16,
            checkpoint=HubFile.Config(
                repo_id='MiniMaxAI/VTP-Small-f16d64',
                filename='model.safetensors',
                revision='378967941e66f7f09a9e10218aa8710a930cc635',
                sha256='5442083e078aadc65e51c7893d10177e360139866d4e09ab955b1c7b563eeaee'
            )
        ),
        image_size=256,
        pixel_mean=(0.485, 0.456, 0.406),
        pixel_std=(0.229, 0.224, 0.225),
        dtype_autocast=None,
        latent_norm=ChannelLatentStats.Config(
            stats=UrlFile.Config(
                url='https://raw.githubusercontent.com/MiniMax-AI/VTP/5ce1eb67010fff3c1eed483352483be6a1838556/generation/latent_stats/vtp_s/latents_stats.pt',
                sha256='9f4bb851a6226843c0bdb9b25330faa2a8428e5c10a77f79a974f145cf848eff'
            ),
            multiplier=1.0
        )
    )
