-
Normalize v3 audit schema across JSONL parts; validate datasets loader
Metadata-only compatibility fix. All 10 training conversations, Python code, tests and token counts are unchanged. Legacy data unchanged. -
Add verified Python TACO v3 pilot part 5/5 (2 rows)
Pinned-source tests plus independent oracles; synthetic reviewed solution summaries; exact Qwen2.5-Coder full-chat token counts. Legacy data preserved. -
Add verified Python TACO v3 pilot part 4/5 (2 rows)
Pinned-source tests plus independent oracles; synthetic reviewed solution summaries; exact Qwen2.5-Coder full-chat token counts. Legacy data preserved. -
Add verified Python TACO v3 pilot part 3/5 (2 rows)
Pinned-source tests plus independent oracles; synthetic reviewed solution summaries; exact Qwen2.5-Coder full-chat token counts. Legacy data preserved. -
Add verified Python TACO v3 pilot part 2/5 (2 rows)
Pinned-source tests plus independent oracles; synthetic reviewed solution summaries; exact Qwen2.5-Coder full-chat token counts. Legacy data preserved. -
Add verified Python TACO v3 pilot part 1/5 (2 rows)
Pinned-source tests plus independent oracles; synthetic reviewed solution summaries; exact Qwen2.5-Coder full-chat token counts. Legacy data preserved.