release.json

{
  "created_at": "2026-10-01T15:30:46.614515+00:00",
  "bucket_id": "HuggingFaceBio/Carbon-A-training-data",
  "visibility": "private",
  "model_id": "HuggingFaceBio/GENERanno-eukaryote-1.2b-cds-annotator",
  "tokenizer_revision": "49f69ea2085657034de291e02cb7791693c6872b",
  "evaluation_revision": "617613c8daa62345d79d0208b76b96149f6604af",
  "train_files": 4561,
  "test_files": 26,
  "train_bytes": 10155092707328,
  "test_bytes": 22126579712,
  "context_tokens": 16384,
  "bases_per_token": 6,
  "bases_per_strand": 98304,
  "label_transform": "nonzero to 1; zero to 0; concatenate plus then minus; special-token positions masked to -100",
  "evaluation_scope": "Original 26-genome evaluation panel used during training; not a new untouched test split.",
  "changes_to_arrays": "None; server-side copies by content hash.",
  "schema_inspection": "Headers and first rows sampled from each of the five training folders and one evaluation file; not a full corpus content audit."
}