{
  "schema": "lewm-browser-export/1",
  "graphVersion": 1,
  "opset": 18,
  "torchVersion": "2.12.0",
  "checkpoint": {
    "repoId": "quentinll/lewm-tworooms",
    "revision": "77adaae0bc31deab21c93740d1f8bb947cd0bdec",
    "weightsSha256": "566f223624ea4bfb39dbfe6ae731198dd6ea73b7b8919fed6b1ecafca810f7dd"
  },
  "architecture": {
    "hiddenDim": 192,
    "imageSize": 224,
    "patchSize": 14,
    "numFrames": 3,
    "actionDim": 10
  },
  "files": {
    "lewm_tworooms_encoder.onnx": {
      "sha256": "736777c330ab3df2a080810b1b781124ec5ded16d5f1f235ba76b538d7122d81",
      "bytes": 25923986,
      "graph": "encoder",
      "inputs": {
        "pixels": [
          "batch",
          3,
          "image_size",
          "image_size"
        ]
      },
      "inputUnits": "RGB floats in [0,1]; ImageNet normalization applied inside the graph",
      "outputs": {
        "latent": [
          "batch",
          "hidden_dim"
        ]
      },
      "components": [
        "imagenet-normalize",
        "encoder (HF-ViT)",
        "projector (Linear-BN-GELU-Linear)"
      ]
    },
    "lewm_tworooms_action.onnx": {
      "sha256": "3d936cf45df0a168934e50a75aae24fc0ff4e8a39ddbe883cf23777faed5904e",
      "bytes": 641851,
      "graph": "action_encoder",
      "inputs": {
        "actions": [
          "batch",
          "time",
          "action_dim"
        ]
      },
      "inputUnits": "raw env action blocks (frameskip-5 x 2D, each in [-1,1]); expert-dataset z-score applied inside the graph",
      "outputs": {
        "action_embedding": [
          "batch",
          "time",
          "hidden_dim"
        ]
      },
      "components": [
        "zscore-normalize",
        "action_encoder (Conv1d-SiLU-MLP)"
      ]
    },
    "lewm_tworooms_predictor.onnx": {
      "sha256": "42d0cb06ab1775d15a81280f42f3d0b3fdc9b2fc4eae619f92c231b5d9f3419e",
      "bytes": 46910027,
      "graph": "predictor",
      "inputs": {
        "latents": [
          "batch",
          "time<=num_frames",
          "hidden_dim"
        ],
        "action_embeddings": [
          "batch",
          "time<=num_frames",
          "hidden_dim"
        ]
      },
      "outputs": {
        "predicted_latents": [
          "batch",
          "time<=num_frames",
          "hidden_dim"
        ]
      },
      "components": [
        "predictor (AdaLN causal transformer)",
        "pred_proj (Linear-BN-GELU-Linear)"
      ]
    }
  },
  "totalBytes": 73475864,
  "normalization": {
    "pixels": {
      "where": "inside lewm_tworooms_encoder.onnx",
      "kind": "imagenet",
      "mean": [
        0.485,
        0.456,
        0.406
      ],
      "std": [
        0.229,
        0.224,
        0.225
      ],
      "input": "RGB floats in [0,1], CHW"
    },
    "actions": {
      "where": "inside lewm_tworooms_action.onnx",
      "kind": "expert-dataset z-score (per 2D dim, tiled over the frameskip-5 block)",
      "input": "raw env actions in [-1,1]",
      "stats": {
        "schema": "lewm-action-stats/1",
        "source": "docs/evidence/lewm_tworooms_action_stats.json",
        "dataset": "quentinll/lewm-tworooms",
        "datasetRevision": "6903a2de048b13819d812da0b4dd661290bc01e4",
        "mean": [
          0.00309341,
          -0.05298233
        ],
        "std": [
          0.86747581,
          0.86776555
        ]
      }
    }
  },
  "runtime": {
    "preferred": "onnxruntime-web webgpu",
    "fallback": "onnxruntime-web wasm",
    "unsupported": "participants report an explicit unsupported state; no silent surrogate fallback in the real mode"
  },
  "parity": {
    "onnxruntimeAvailable": true,
    "status": "passed",
    "atol": 0.0001,
    "maxAbsDiff": 1.0013580322265625e-05,
    "checks": [
      {
        "graph": "lewm_tworooms_encoder.onnx",
        "inputShapes": [
          [
            2,
            3,
            224,
            224
          ]
        ],
        "maxAbsDiff": 1.0013580322265625e-05
      },
      {
        "graph": "lewm_tworooms_action.onnx",
        "inputShapes": [
          [
            2,
            3,
            10
          ]
        ],
        "maxAbsDiff": 1.6689300537109375e-06
      },
      {
        "graph": "lewm_tworooms_action.onnx",
        "inputShapes": [
          [
            1,
            7,
            10
          ]
        ],
        "maxAbsDiff": 9.5367431640625e-07
      },
      {
        "graph": "lewm_tworooms_predictor.onnx",
        "inputShapes": [
          [
            2,
            3,
            192
          ],
          [
            2,
            3,
            192
          ]
        ],
        "maxAbsDiff": 1.3113021850585938e-06
      },
      {
        "graph": "lewm_tworooms_predictor.onnx",
        "inputShapes": [
          [
            1,
            1,
            192
          ],
          [
            1,
            1,
            192
          ]
        ],
        "maxAbsDiff": 4.6193599700927734e-07
      },
      {
        "graph": "lewm_tworooms_predictor.onnx",
        "inputShapes": [
          [
            1,
            2,
            192
          ],
          [
            1,
            2,
            192
          ]
        ],
        "maxAbsDiff": 1.8477439880371094e-06
      }
    ]
  },
  "nonClaims": [
    "Checkpoint-backed LeWorldModel inference graphs for the Tapestry-like demo; not browser training artifacts, not a paper-scale benchmark claim."
  ]
}