{
  "scope": "Genuine original V-JEPA 2 ViT-L frozen features; each photograph repeated for 16 frames; appearance only, no motion or action recognition.",
  "provenance": {
    "architecture": "Original V-JEPA 2 ViT-L/16, 24 layers, width 1024, 16 heads, 2-frame tubelets",
    "revision": "204698b45b3712590f06245fbfba32d3be539812",
    "checkpointUrl": "https://dl.fbaipublicfiles.com/vjepa2/vitl.pt",
    "checkpointSha256": "5346856ec9df69487fe72a25bf2632aaa8112df33fb67708e3f7374edc1f7012",
    "checkpointKey": "target_encoder",
    "removedPrefix": "module.backbone.",
    "strictLoadedTensors": 292,
    "ignoredTensors": [],
    "sourceHashes": {
      "src/models/vision_transformer.py": "9d4fd4e3603311803c105cc5114e2bc12ff28bc11870637dd3969511d9dead74",
      "src/models/utils/modules.py": "b93f6c7e0747deb216419c000c2878f11a9189024a9adeacfd437e172396dff0",
      "src/models/utils/patch_embed.py": "1823c40b34410e70eb313360e21678e18a0ee5465b5d47de9f3d967c074f5a9d",
      "src/models/utils/pos_embs.py": "e6d3b3863ade47b3734a3b443f1bcc93eab02e39bf596ea199517f68ccab443e",
      "src/masks/utils.py": "833f111a0fa5ffdbd3a6412e2dace2517c3c178f49c14f8bb631d9f6a070dfd0",
      "src/utils/tensors.py": "782b58bd2af456e184750e5318ab773105108383f61b280fe4c7a90f46add2c8"
    },
    "programSha256": "0612a63eda5f3b8d85e2725b7bfc027f15580dd987985d928f7cc140e188ab7b",
    "parameterCount": 303885312,
    "python": "3.13.13",
    "torch": "2.12.0+cpu",
    "numpy": "2.1.2",
    "pillow": "12.2.0",
    "device": "cpu",
    "dtype": "float32",
    "command": "python examples/vjepa2_export.py --checkpoint .context/paper-series/weights/vjepa2-vitl.pt --official-dir .context/paper-series/research/vjepa2-native --photos-dir public/examples/vit-demo --output public/examples/vjepa2-demo",
    "adapter": "Hash-pinned native author encoder; native timm drop_path replacement; unused einops import removed; released RoPE frequency-repeat behavior preserved."
  },
  "preprocessing": {
    "resize": "PIL bilinear, shortest edge 256; floor long edge",
    "crop": "256x256 center crop; floor offsets",
    "frames": 16,
    "temporalInput": "16 identical copies of a still photograph, no measured motion",
    "mean": [
      0.485,
      0.456,
      0.406
    ],
    "std": [
      0.229,
      0.224,
      0.225
    ],
    "pooling": "Mean of all 2048 final-layer patch features, without extra feature normalization"
  },
  "photos": [
    {
      "id": "cat",
      "title": "Chelsea the cat",
      "source": {
        "url": "https://raw.githubusercontent.com/scikit-image/scikit-image/v0.25.2/skimage/data/chelsea.png",
        "sha256": "596aa1e7cb875eb79f437e310381d26b338a81c2da23439704a73c4651e8c4bb",
        "author": "Stefan van der Walt",
        "license": "CC0",
        "licenseUrl": "https://github.com/scikit-image/scikit-image/blob/v0.25.2/skimage/data/_fetchers.py"
      },
      "inputImage": "input-cat.png",
      "inputSha256": "e88909c9e7d274881fc00b9b5733ea3aaa0e9ea4f0a033c28959e11f9efe34e0",
      "inputShape": [
        1,
        3,
        16,
        256,
        256
      ],
      "resizedSize": [
        384,
        256
      ],
      "cropBox": [
        64,
        0,
        320,
        256
      ],
      "tokenShape": [
        1,
        2048,
        1024
      ],
      "meanTokenL2": 104.81875610351562,
      "pooledL2": 53.193626403808594,
      "pooledFirst8": [
        0.14168523252010345,
        -0.8578346967697144,
        0.012752451002597809,
        -2.1759228706359863,
        -2.64709210395813,
        -0.06411787867546082,
        0.39415913820266724,
        -0.7345725297927856
      ]
    },
    {
      "id": "coffee",
      "title": "Cup of coffee",
      "source": {
        "url": "https://raw.githubusercontent.com/scikit-image/scikit-image/v0.25.2/skimage/data/coffee.png",
        "sha256": "cc02f8ca188b167c775a7101b5d767d1e71792cf762c33d6fa15a4599b5a8de7",
        "author": "Rachel Michetti",
        "license": "CC0",
        "licenseUrl": "https://github.com/scikit-image/scikit-image/blob/v0.25.2/skimage/data/_fetchers.py"
      },
      "inputImage": "input-coffee.png",
      "inputSha256": "8e202cf8a944b65408ffb34e1c1a21fdadf17bdc35da84405ea4cb1d514a5716",
      "inputShape": [
        1,
        3,
        16,
        256,
        256
      ],
      "resizedSize": [
        384,
        256
      ],
      "cropBox": [
        64,
        0,
        320,
        256
      ],
      "tokenShape": [
        1,
        2048,
        1024
      ],
      "meanTokenL2": 104.70674133300781,
      "pooledL2": 50.43646240234375,
      "pooledFirst8": [
        0.02447730302810669,
        -0.1830258071422577,
        -0.41385945677757263,
        0.08480307459831238,
        -4.158973217010498,
        0.058460451662540436,
        0.2781641185283661,
        0.4911375939846039
      ]
    }
  ],
  "cosineSimilarity": [
    [
      0.9999999403953552,
      0.5525920391082764
    ],
    [
      0.5525920391082764,
      1.000000238418579
    ]
  ],
  "features": {
    "file": "pooled-features.npy",
    "shape": [
      2,
      1024
    ],
    "sha256": "128699e79187f613e860ab783c8bf840239acb0791f40ae5fdd27224e78f8dbe"
  }
}
