Hacktoberfest 2026:维护者为十月标记出来的 issue,仍然开放、适合新手。 浏览 Hacktoberfest issue

[Wave 4] Low/info — TR mirror: docs/usermanuals/en/concepts/

未关闭 适合新手
#812 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
2/5
预计耗时
1-3 小时
新手友好度
62/100
Issue 类型
文档
描述清晰度
基本清楚
活跃度
活跃
技术栈
python
领域
documentation

调研方向

从土耳其语镜像页面 docs/usermanuals/en/concepts/alignment-overview-tr.md 和 data-formats-tr.md 开始,并与 issue 中引用行号处的英文原文进行比较。修正 alignment-overview 的标题(EN 第 117 行,TR 第 117 行),恢复 data-formats 中的完成项列表(TR 第 88-90 行),并将 reward() 示例(TR 第 105-112 行)替换为受支持的数据集列流程。当三项发现都已在土耳其语中修正并在清单中勾选时,即视为完成。

由索引模型根据 Issue 内容生成。

描述

documentation i18n severity: low source: review-2026-10 wave: 4

Wave 4 · 3 finding(s) in 3 fix package(s). Mirror drift across small document pairs in this directory.

Findings

  • DOC-TR13-003 (TR mirror, Low) — TR mirror drift in docs/usermanuals/en/concepts/alignment-overview.md: ## Where to go next
    severity Low
    Where. EN alignment-overview.md:117-121 (heading 'Where to go next') · TR alignment-overview.md:117 ('## Sıraki adım')
    English text. ## Where to go next
    Turkish text. ## Sıraki adım
    Which is correct. EN (TR has a typo)
    Proposed correction (Turkish text for the -tr page). ## Sıradaki adım

  • DOC-TR13-004 (TR mirror, Low) — TR mirror drift in docs/usermanuals/en/concepts/data-formats.md: completion → the model's response (the field name TRL's KTOTrainer expects; ForgeLM's _detect_dataset_format keys off completion… severity **Low** **Where.** EN [data-formats.md:89](https://github.com/HodeTech/ForgeLM/blob/cab2947/data-formats.md#L89) (completionfield meaning +_detect_dataset_format keys off completion+label to auto-route to KTO) · TR [data-formats.md:88-90](https://github.com/HodeTech/ForgeLM/blob/cab2947/data-formats.md#L88-L90) (only label meanings; the completionbullet is absent) **English text.**completion→ the model's response (the field name TRL's KTOTrainer expects; ForgeLM's_detect_dataset_formatkeys offcompletion + labelto auto-route to KTO). **Turkish text.** (bullet missing; list starts at 'label: true → istenen yanıt') **Which is correct.** TR omission is harmless: the dropped sentence is the inaccurate one (detection is advisory only, data.py:23-50). Thecompletionfield meaning is still useful. **Proposed correction (Turkish text for the -tr page).** -completion→ modelin yanıtı (TRL KTOTrainer'ın beklediği alan adı; sütunlarcompletion+labeliçermezse ForgeLMKeyError` verir, formatı otomatik yönlendirmez).

  • DOC-TR13-005 (TR mirror, Low) — TR mirror drift in docs/usermanuals/en/concepts/data-formats.md: """Return a scalar in [-1, 1] (or any bounded range).""" and comments '# your parsing logic', '# malformed'
    severity Low
    Where. EN data-formats.md:106-114 (reward() docstring 'Return a scalar in [-1, 1] (or any bounded range)') · TR data-formats.md:105-112 (docstring absent)
    English text. """Return a scalar in [-1, 1] (or any bounded range).""" and comments '# your parsing logic', '# malformed'
    Turkish text. (docstring and inline comments removed)
    Which is correct. Neither: the whole reward(prompt,response,ground_truth) example is not wired in ForgeLM (see D-0940); TR drops only comments.
    Proposed correction. Replace the example in both languages with the supported flow: dataset column gold_answer + leave grpo_reward_model null, or an HF reward model id.

Done when

  • Every finding above is fixed or closed with a reason (tick it here).
  • Code fixes carry a test; doc fixes update the Turkish mirror in the same PR.

Found by the October 2026 ForgeLM code and documentation review against commit cab2947. Code links point at that commit; line numbers may have moved since. Finding IDs (CR-…, DOC-…, TECH-…) are stable references for this backlog.

主要语言
Python
星标
9
派生
1
平均合并
8 小时 47 分钟
30 天内合并 PR
47

环境准备

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

HodeTech/ForgeLM 的其他 Issue

查看 HodeTech/ForgeLM 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。