five

code2lora/code2lora-data-ood

收藏
Hugging Face2026-05-25 更新2026-05-31 收录
下载链接:
https://hf-mirror.com/datasets/code2lora/code2lora-data-ood
下载链接
链接失效反馈
官方服务:
资源简介:
--- license: mit tags: [code, assertion-completion, repository-level, ood] --- # code2lora-data-ood — post-cutoff OOD evaluation split Out-of-distribution evaluation slice: commits from repositories whose first release is **strictly after** the Qwen2.5-Coder pretraining cutoff. Used to isolate cross-repository transfer from training-set memorisation. ## Layout ``` commits.parquet per-commit metadata + diff/repo-state embeddings qna.parquet per-assertion rows (prefix, target, lineno, ...) ``` Both files share the same `(repo_id, commit_index, commit_sha)` key, so a left join on those three columns reconstructs the per-commit QnA bundle. ## Schema `commits.parquet` matches `code2lora/code2lora-data-commits/commits/*.parquet`. `qna.parquet` matches `code2lora/code2lora-data-snapshots/qna/*.parquet`. ## Companion datasets * `code2lora/code2lora-data-commits` in-distribution per-commit dataset * `code2lora/code2lora-data-snapshots` static-snapshot variant
提供机构:
code2lora
二维码
社区交流群
二维码
科研交流群
商业服务