{
  "native/__init__.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/__init__.py",
    "original_sha256": "61cc47f51b054b44b0e2d42e845e52e77941dcc52e0818b9e3b8b99e7d11b12f",
    "modifications": []
  },
  "native/configuration_rqvaesigliptransformer.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/configuration_rqvaesigliptransformer.py",
    "original_sha256": "e25b545dffef9f456399b20dfc7e03667221f5960320c22786a21a5267db4865",
    "modifications": []
  },
  "native/modeling_rqvaesigliptransformer.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/modeling_rqvaesigliptransformer.py",
    "original_sha256": "f3f0ea61252e376d2e9ca3d8ce64bd93c2e3f88847e345fc4e0f2cdebf3b4f3b",
    "modifications": []
  },
  "native/rqtransformer/__init__.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqtransformer/__init__.py",
    "original_sha256": "64a367f014a6c68d30e1778df60d44dcd73f4bb700d5f0811f9b1f55fdf77511",
    "modifications": []
  },
  "native/rqtransformer/attention.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqtransformer/attention.py",
    "original_sha256": "335c91a006801207183106161b93b1b3c3eed00c38e71f1579cbe084ac189b88",
    "modifications": []
  },
  "native/rqtransformer/configuration_rqtransformer.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqtransformer/configuration_rqtransformer.py",
    "original_sha256": "93d63066940b5f8837aa9fccdf3fc4c719d99ec1d9b4f54db9dffacb0587e86c",
    "modifications": [
      "Replace the mutable dataclass default with default_factory for Python 3.11+; field values and tensor computation unchanged."
    ]
  },
  "native/rqtransformer/modeling_rqtransformer.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqtransformer/modeling_rqtransformer.py",
    "original_sha256": "9b0016b405266b407f63e99649a1dfa8ad2363804bb6a27bd6de03643cf28fcf",
    "modifications": []
  },
  "native/rqvaesiglip/__init__.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/__init__.py",
    "original_sha256": "45ea5844cfcb8256ca1f2cff9af333dd30412baa6782aa00d8fe9ee1b15562b2",
    "modifications": []
  },
  "native/rqvaesiglip/configuration_rqvaesiglip.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/configuration_rqvaesiglip.py",
    "original_sha256": "0974d7090d62870b1a1e1fa507fd8500c636b8918ab33a84998962255a928c2a",
    "modifications": [
      "Add an optional bundled SigLIP architecture dictionary to eliminate runtime configuration downloads; tensor computation and trained configuration values unchanged."
    ]
  },
  "native/rqvaesiglip/modeling_rqvaesiglip.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/modeling_rqvaesiglip.py",
    "original_sha256": "5d2a2a6ca6dc08f3b87790a409b2469a3fe8ba7fc60af1b5432d9105ba21fc14",
    "modifications": [
      "Read the same pinned SigLIP architecture config from the bundled dictionary instead of fetching it at model construction; no computation changes.",
      "Optional decoder latent resizing similarly preserves input dtype instead of hardcoded bfloat16."
    ]
  },
  "native/rqvaesiglip/modules.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/modules.py",
    "original_sha256": "44fc573058dc39f9488f90f471cd1ce29d77994c6abbfb4bfc0a3a33c535803f",
    "modifications": [
      "After float32 nearest interpolation, cast back to the input dtype rather than hardcoded bfloat16. Original bfloat16 output arithmetic is retained and tested."
    ]
  },
  "native/rqvaesiglip/quantizations.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/quantizations.py",
    "original_sha256": "b71468ce78ed1e773f8e491734a05b8ae9703498c0cda097a533c159605b4a13",
    "modifications": []
  },
  "native/rqvaesiglip/siglip/__init__.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/siglip/__init__.py",
    "original_sha256": "83892d51629e08b95e91e33e10b2df066422878521ec339f8be73c570181ad6e",
    "modifications": []
  },
  "native/rqvaesiglip/siglip/configuration_siglip.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/siglip/configuration_siglip.py",
    "original_sha256": "cf274db690f28f81c359e0b6fe2ead75a1e88386595dc40df4e70ff413465b3f",
    "modifications": []
  },
  "native/rqvaesiglip/siglip/convert_siglip_to_hf.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/siglip/convert_siglip_to_hf.py",
    "original_sha256": "460e418515957882b173d0f3d195028e11b785ab0db6b39194870e62857bc52d",
    "modifications": []
  },
  "native/rqvaesiglip/siglip/image_processing_siglip.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/siglip/image_processing_siglip.py",
    "original_sha256": "9d19ce5a8710b138a08fe1b7730fa3c401b7df08692924fe52d2d81fc6605a21",
    "modifications": []
  },
  "native/rqvaesiglip/siglip/modeling_siglip.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/siglip/modeling_siglip.py",
    "original_sha256": "12111cb8156112b2ad356d7ec3127656d1b3f11365e5d4361c4fa7077de17a55",
    "modifications": []
  },
  "native/rqvaesiglip/siglip/processing_siglip.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/siglip/processing_siglip.py",
    "original_sha256": "5b9fadb4b6f82d3fcf1f58f7c4d5613ad7a4a23f17fe158218b2145aa6e02c05",
    "modifications": []
  },
  "native/rqvaesiglip/siglip/tokenization_siglip.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_encoder/rqvaesigliptransformer/rqvaesiglip/siglip/tokenization_siglip.py",
    "original_sha256": "50d9570b76e601e88bfc17e962df9c8f81e9d2e83eaf8e24a730c381947d7ac4",
    "modifications": []
  },
  "LICENSE": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/LICENSE",
    "original_sha256": "480a5f7ac02bb268cbc83cbfdaf6d459144d66cfc68262fb969da8113067d955",
    "modifications": []
  },
  "siglip_config.json": {
    "url": "https://huggingface.co/google/siglip-large-patch16-256/resolve/d0da9f876e7d66b4e250cd2450c3ba2ce735e447/config.json",
    "original_sha256": "1053d489ef9053689fc5ee3659c2212c6a44a6f0d71daeabb23a163a3150e57d",
    "modifications": [],
    "note": "Architecture metadata only. All trained SigLIP weights remain the original VILA-U checkpoint values."
  },
  "RQ_VAE_LICENSE": {
    "url": "https://raw.githubusercontent.com/kakaobrain/rq-vae-transformer/main/LICENSE",
    "original_sha256": "3d56f1ff69ee804d90f5c8fe928aa4099afd2d00e5c66b3dfca85f00198e48d2",
    "modifications": [],
    "purpose": "Original third-party license/attribution, alongside the main model/project terms."
  },
  "APACHE_LICENSE": {
    "url": "https://www.apache.org/licenses/LICENSE-2.0.txt",
    "original_sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30",
    "modifications": [],
    "purpose": "Original third-party license/attribution, alongside the main model/project terms."
  },
  "source_projector.py": {
    "url": "https://raw.githubusercontent.com/mit-han-lab/vila-u/4ea42ead2ce22b035ee74c7966eed245bca3e927/vila_u/model/multimodal_projector/base_projector.py",
    "original_sha256": "21ae4db93a001b0e9f086628295dce9ae19f0a53cbe0cee547a9c7a0696ed135",
    "modifications": [],
    "purpose": "Native projector constructor reference. The standalone wrapper constructs its exact configured MLP without importing unrelated model builders."
  },
  "preprocessor_config.json": {
    "url": "https://huggingface.co/mit-han-lab/vila-u-7b-256/resolve/91f97e4b4746e6c2a27a168726ff5a2b9fe0d850/vision_tower/preprocessor_config.json",
    "original_sha256": "215f74bded354b037bd29589109c68cbde8b4e5ca72ebe2b6e02e7c84b7b6856",
    "modifications": []
  }
}