VTP.Config(
        encoder=Encoder.Config(
            patch_size=16,
            channels_hidden=768,
            num_layers=12,
            heads=12,
            expansion=4.0,
            channels_out=64,
            checkpoint=HubFile.Config(
                repo_id='MiniMaxAI/VTP-Base-f16d64',
                filename='model.safetensors',
                revision='21fe3a9cb53f01a2c7f3b622efdc8d2534a7c79d',
                sha256='c7fdbb1507cecbcbb35eab4ea9fcbca8dc80b5e0d64cfe0c37b7baacdbb0fa05'
            )
        ),
        decoder=Decoder.Config(
            channels_in=64,
            channels_hidden=768,
            num_layers=12,
            heads=12,
            expansion=4.0,
            upscale_factor=16,
            checkpoint=HubFile.Config(
                repo_id='MiniMaxAI/VTP-Base-f16d64',
                filename='model.safetensors',
                revision='21fe3a9cb53f01a2c7f3b622efdc8d2534a7c79d',
                sha256='c7fdbb1507cecbcbb35eab4ea9fcbca8dc80b5e0d64cfe0c37b7baacdbb0fa05'
            )
        ),
        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_b/latents_stats.pt',
                sha256='f7c030310f3eeaa38d2d0838f38eeab1d9922fcaf183a0fe8dd1f1287366c4a7'
            ),
            multiplier=1.0
        )
    )
