RAE.Config(
        encoder=Encoder.Config(
            channels_hidden=768,
            num_layers=12,
            heads=12,
            expansion=4,
            patch_size=14,
            image_size=518,
            num_register_tokens=4,
            eps=1e-06,
            activation=<built-in function gelu>,
            checkpoint=HubFile.Config(
                repo_id='facebook/dinov2-with-registers-base',
                filename='model.safetensors',
                revision='a1d738ccfa7ae170945f210395d99dde8adb1805',
                sha256='7a6f7b3b9fa4b8732e707476a03cd6cdce210048582f21aafb7991c17d98e362'
            )
        ),
        decoder=Decoder.Config(
            channels_in=768,
            num_patches=256,
            channels_hidden=1_152,
            channels_hidden_mlp=4_096,
            num_layers=28,
            heads=16,
            patch_size=16,
            eps=1e-12,
            activation=<built-in function gelu>,
            checkpoint=HubFile.Config(
                repo_id='nyu-visionx/RAE-collections',
                filename='decoders/dinov2/wReg_base/ViTXL_n08/model.pt',
                revision='1be4f03273523431f099a934da4cf1940dc6039f',
                sha256='5fedf7c9660476a709e122cef18385c917532914a23f034388e1c1a52bde2be6'
            )
        ),
        image_size=256,
        encoder_image_size=224,
        pixel_mean=(0.485, 0.456, 0.406),
        pixel_std=(0.229, 0.224, 0.225),
        latent_norm=ElementwiseLatentStats.Config(
            stats=HubFile.Config(
                repo_id='nyu-visionx/RAE-collections',
                filename='stats/dinov2/wReg_base/imagenet1k/stat.pt',
                revision='1be4f03273523431f099a934da4cf1940dc6039f',
                sha256='84ede66def5e6e3f25679334dc89cf63b12aacb99cbf0f5ae7ed4ad3187f7e59'
            ),
            eps=1e-05
        )
    )
