Skip to content

docs: ground punctuation tutorial in observed CPU output - #3693

Merged
LauraGPT merged 1 commit into
mainfrom
codex/punctuation-reader-20260910
Sep 10, 2026
Merged

docs: ground punctuation tutorial in observed CPU output#3693
LauraGPT merged 1 commit into
mainfrom
codex/punctuation-reader-20260910

Conversation

@LauraGPT

Copy link
Copy Markdown
Collaborator

Summary

Rewrite the existing Chinese and English punctuation tutorial around one practical workflow: retain the original, generate a separate candidate, and review meaning before publishing.

Show the actual output of the published FunASR 1.4.15 CPU package, including an incorrect boundary and the English leading space. Label the human edit separately. Explain the standalone ct-punc case/whitespace and final-period rules with pinned source references, without extrapolating them to ASR/VAD reconstruction or native Transformers Nano.

Keep the URLs and original publication date. Mark this existing explanation reviewed; keep the homepage at one lead and four selected stories. No runtime change, issue closure, accuracy claim or PyPI release.

Validation

  • 511 final product-site tests passed.
  • Six final browser cases passed: EN/ZH at 320/390/1440px; discovery, language switching, overflow and code-copy controls.
  • Exact bilingual article code executed on the installed FunASR 1.4.15 wheel with PyTorch/torchaudio 2.10.0+cpu. All three JSON candidates match byte-for-byte, including whitespace; pip check passed.
  • Installed ct_transformer code matches the pinned source. Model files and actual weight SHA256 recorded; mutable ModelScope master is explicitly not an immutable revision.
  • Two identical builds: 204 HTML pages and 271 files. Only the two articles, four category/archive indexes and two search indexes differ; homepage, native Nano article and model pages are unchanged.
  • Independent content/source review found no P1/P2. Six representative screenshots inspected. Post-review changes only remove two EOF blank lines from tests; all final site/browser gates rerun.
  • Before-push verified bundle and explicit-file backup preserved. Signed commit and DCO.

Retained diagnostic history: nine article assertions failed before editing. MacOS archive metadata caused one build failure and was quarantined after byte/untracked verification; no builder change. The first full-suite command used a venv without xdist and failed before tests; the existing system-Python test environment passed. This article does not claim to resolve FunASR #3644.

Signed-off-by: zhifu gao <zhifu.gzf@alibaba-inc.com>
@LauraGPT
LauraGPT merged commit 535e0d9 into main Sep 10, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant