docs(links): fix broken documentation links - #311
Open
ooooo-create wants to merge 5 commits into
Open
Conversation
# 📝 Documentation - Correct DeepEyes dataset and Megatron Bridge links in English and Chinese docs. - Fix same-language autoscaler links to elastic rollout and fully async training. - Point math OPD examples to the current script and align model and configuration descriptions. ## Validation - Run all pre-commit hooks successfully. - Verify changed relative link targets exist. - Skip multi-node GPU integration tests because only documentation changed.
# 📝 Documentation - Correct English and Chinese low-precision training links to match existing heading anchors. - Point the MOPD configuration reference to the OPD common combinations section. - Preserve heading text without adding explicit anchor IDs. ## Validation - Pass Lychee offline fragment checks: 8 links, 0 errors. - Pass all pre-commit hooks and staged diff checks. - Skip multi-node GPU integration tests because only documentation links changed.
ooooo-create
requested review from
Aurelius84,
NINGBENZHE,
Yangruipis and
yxyOo
as code owners
September 11, 2026 14:27
# 📝 Documentation - Remove the added flag-mapping explanation from Common Combinations. - Preserve the original wording and corrected script link. ## Validation - All pre-commit hooks and staged diff checks pass.
# 📝 Documentation - Restore the original student_sampled and adv mode description. - Keep the corrected script link and updated student/teacher model names. ## Validation - All pre-commit hooks and staged diff checks pass.
rai-studio-bot
left a comment
rai-studio-bot
left a comment
There was a problem hiding this comment.
已复查最新提交并补全全部文档差异检查，未发现需要阻塞合入的问题。低精度文档的章节链接仍有 VitePress 锚点兼容问题，详情见行内审查意见。
CI 检查均已通过。本次未运行完整 VitePress 构建；仅涉及文档修改，未运行多节点 GPU 集成测试。
Powered by Nyanpasu with gpt-6-astra medium, please check the suggestions carefully.
| 2. The Megatron patch at `docker/patch/megatron/20260506-85bced0ae.patch` applied (baked into the project Dockerfile). It provides both the FP8 overrides and the INT4 `_FakeInt4QuantizationSTE` that overrides `TEGroupedLinear._get_weight_tensors()`. | ||
|
|
||
| The FP8 recipe additionally needs a TransformerEngine build with FP8 blockwise scaling support. The INT4 recipe additionally needs the `fake_int4_quant_cuda` CUDA extension built — see [Build the int4_qat kernel](#build-the-int4-qat-kernel) below. | ||
| The FP8 recipe additionally needs a TransformerEngine build with FP8 blockwise scaling support. The INT4 recipe additionally needs the `fake_int4_quant_cuda` CUDA extension built — see [Build the int4_qat kernel](#build-the-int4_qat-kernel) below. |
There was a problem hiding this comment.
请按 VitePress 的标题 ID 修正这组锚点。项目使用 VitePress，且 .vitepress/config.mts 未覆盖默认 slugify；VitePress 1.6.4 使用的 @mdit-vue/shared 会把下划线和句点转成连字符。实际调用该函数得到 build-the-int4-qat-kernel、编译-int4-qat-kernel 和 convert-hf-to-int4-py。当前改法虽符合 GitHub Markdown 的锚点规则，却会使文档站的编译链接失效，且量化工具链接仍无法跳转；离线 GitHub fragment 检查不能覆盖这种差异。
本行可恢复为：
Suggested change
| The FP8 recipe additionally needs a TransformerEngine build with FP8 blockwise scaling support. The INT4 recipe additionally needs the `fake_int4_quant_cuda` CUDA extension built — see [Build the int4_qat kernel](#build-the-int4_qat-kernel) below. | |
| The FP8 recipe additionally needs a TransformerEngine build with FP8 blockwise scaling support. The INT4 recipe additionally needs the `fake_int4_quant_cuda` CUDA extension built — see [Build the int4_qat kernel](#build-the-int4-qat-kernel) below. |
请同步修正中文页面两处编译链接，以及中英文页末的量化工具链接：
#编译-int4_qat-kernel → #编译-int4-qat-kernel
#convert_hf_to_int4py → #convert-hf-to-int4-py
若需兼容两种渲染器，可另加对应的显式锚点；验证时应检查 VitePress 渲染后的标题 ID。
# 📝 Documentation - Add explicit HTML anchors for HF-to-INT4 conversion and INT4 QAT kernel builds in both languages. - Point all six section references to shared IDs while preserving heading text. - Avoid differences between GitHub Markdown and VitePress heading slug generation. ## Validation - GitHub Markdown rendering retains both anchors on each page. - VitePress 1.6.4 rendering resolves all six references with unique IDs. - Lychee offline fragment checks pass: 8 links, 0 errors. - All pre-commit hooks and staged diff checks pass.
rai-studio-bot
approved these changes
Sep 11, 2026
rai-studio-bot
left a comment
rai-studio-bot
left a comment
There was a problem hiding this comment.
已复查 0b8c0a6，上轮锚点兼容问题已解决：中英文页面改用显式 HTML 锚点，六处章节引用均匹配唯一目标。未发现新的问题，无需新增行内意见。
本次核验了源码中的锚点及引用，未重跑完整文档构建；PR 描述已补充完整构建和渲染验证记录。Pre-commit、Python 3.10/3.12 检查已通过，Python 3.11 检查仍在运行。仅涉及文档修改，未运行多节点 GPU 集成测试。
Powered by Nyanpasu with gpt-6-astra medium, please check the suggestions carefully.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Fix broken dataset, source, script, and section links in the English and Chinese documentation and OPD READMEs.
Why
The existing links point to missing files, an outdated dataset location, or fragments affected by different automatic heading-ID rules in GitHub Markdown and VitePress.
How
Testing
pre-commit run --all-filespassespytest tests/)All pre-commit hooks passed with
--show-diff-on-failure. Changed relative file targets and diff whitespace checks passed. Lychee offline fragment checks passed for both low-precision training pages and the MOPD README. GitHub Markdown rendering retains both explicit anchors on each page; VitePress 1.6.4 rendering confirms unique IDs and valid targets for all six low-precision section references.Python and multi-node GPU integration tests were not run because only documentation changed; new tests are not applicable. The complete
npm run docs:buildpassed with Node.js 22.22.1, Python 3.11.15, and VitePress 1.6.4, including OpenAPI generation, page rendering, and chunk filename post-processing. Both generated low-precision HTML pages were checked: all six section references target unique explicit anchors.Type of Change
Screenshots / Logs