{
  "source_project": "fp_finetune_jsonl",
  "source_file": "chat/all_combined.jsonl",
  "source_sha256": "a67de5f91ce6a0943c532eed637b5b476f397080ec5cad6097a578bdd5acff0a",
  "source_bytes": 1575780,
  "matching_uploaded_copy": {
    "project": "LLM_FINE_TUNE/finetune",
    "file": "data/datasets/all_combined-uploaded-20260520-125234.jsonl",
    "sha256": "827a3d763b2ef3477bcb96bbda8021c70526d78e86a0b92f589db3c4913a6da2",
    "identical_parsed_records_and_order": true,
    "byte_identical": false
  },
  "training_notebook_sha256": "e17b12c5d25b949fdb601b882604cfbb0a6142dcb13faa887b0219f39e44908e",
  "training_notebook_recorded_examples": 3657,
  "training_run_steps": 1374,
  "training_epochs": 3,
  "association": "Filename, exact example count, per-task contents, ordered matching uploaded copy, and notebook/checkpoint agreement support this snapshot. The original Colab run did not record a dataset checksum; exact historical byte identity is not proven.",
  "original_creation_process": "Local task-specific instruction examples and short noisy/typo variants. No generation script or external dataset attribution accompanies this exact chat corpus; hand-authoring versus assisted generation is not established.",
  "source_dates_local": {
    "chat_file": "2026-05-20",
    "training_notebook": "2026-05-21"
  },
  "publication_preparation": "Original chat JSONL bytes copied unchanged. Diagnostic test file also unchanged. No filtering, label correction, deduplication or new train/validation split."
}