ounce100m-code / build /audit_contamination.py

Commit History

28 days ago

  1. Upload build/audit_contamination.py with huggingface_hub

    Verified signature
    7f57bf0
  2. Upload build/audit_contamination.py with huggingface_hub

    Verified signature
    1f792cf
  3. Upload build/audit_contamination.py with huggingface_hub

    Verified signature
    1b5ea42
  4. Upload build/audit_contamination.py with huggingface_hub

    Verified signature
    484f967
  5. audit: vectorised rolling-hash grams + sorted-array lookup; the Python-set version could not finish 1.2B tokens

    Verified signature
    66e6f6e
  6. audit: composite AUDIT_PASSED verdict; missing shards are unmeasured, not clean

    Verified signature
    98e4956
  7. smoke PASSED (4/4 sources per shard, 40/40 boundaries); add contamination audit; raw target 1.23B = 23% headroom

    Verified signature
    18d6af8