Skip to content

Upgrade llama.cpp to b9637 and update GitHub Actions versions#234

Merged
bernardladenthin merged 5 commits into
mainfrom
claude/sweet-lamport-ugvqea
Jun 14, 2026
Merged

Upgrade llama.cpp to b9637 and update GitHub Actions versions#234
bernardladenthin merged 5 commits into
mainfrom
claude/sweet-lamport-ugvqea

Conversation

@bernardladenthin

Copy link
Copy Markdown
Owner

Summary

  • Upgrade llama.cpp from b9621 to b9637, which includes Cohere2 MoE architecture support, Jinja chat-template engine fixes, and Vulkan backend improvements
  • Update GitHub Actions to use major version pinning (e.g., v6 instead of v6.0.3) for actions/checkout and actions/setup-java across all workflows
  • Remove -Dmaven.javadoc.skip=true flag from PIT mutation testing command
  • Update documentation to reflect the new llama.cpp version

Test plan

  • CI is green on this branch
  • Docs / CHANGELOG updated where applicable

Related issues / PRs

Checklist

  • I have read CONTRIBUTING.md and CODE_OF_CONDUCT.md
  • My commits follow Conventional Commits
  • No security-sensitive changes (if there are, I have notified the maintainer privately per SECURITY.md)

https://claude.ai/code/session_01EQJCrQGmxCBf8WTCDuFE3X

claude added 5 commits June 14, 2026 18:56
Use @v6.0.3 / @v5.2.0 / @v5.0.5 to match all other workflows (publish.yml,
codeql.yml, scorecard.yml). SHA pins were inconsistent with the rest of the
CI and were introduced by dependabot's default style for this file.

https://claude.ai/code/session_01EQJCrQGmxCBf8WTCDuFE3X
Consistent with publish.yml style. Dependabot will only PR on major bumps,
not on every minor/patch release.

https://claude.ai/code/session_01EQJCrQGmxCBf8WTCDuFE3X
Replace every actions/checkout@v6.0.3 and actions/setup-java@v5.2.0 with
@v6 and @v5 to match the sibling repos and avoid minor/patch dependabot churn.

https://claude.ai/code/session_01EQJCrQGmxCBf8WTCDuFE3X
The flag sat on a `test-compile … mutationCoverage` command, which never reaches
the `package` phase that javadoc:jar binds to — so it skipped nothing. Part of a
cross-repo javadoc-skip cleanup; this repo is Java-8 (<source>1.8>) so javadoc
stays in classpath mode and needs no skip anywhere.

https://claude.ai/code/session_01EQJCrQGmxCBf8WTCDuFE3X
No breaking API changes: none of the project's include surface (common.h,
chat.h, speculative.h, mtmd.h, llama-cpp.h, arg.h, llama.h, download.h) is
touched. The upgrade is purely additive.

New capabilities gained automatically (no project code needed):
- Cohere2 MoE ("North Code") model arch (MoE + MTP/NextN) with a dedicated
  chat parser, auto-detected via the existing specialized-template path.
- Jinja chat-template engine fixes (count/d/e filter aliases, negative-step
  slicing, empty-separator split guard, empty-old_str replace).

Vulkan unary-shader consolidation + EXPM1, WebUI gzip serving, CLI/Docker/CI/
Python-converter changes are all in TUs the project does not compile or ship.

Verified: CMake configures cleanly against b9637 (ggml 0.15.1, CPU backend).
docs/history/llama-cpp-breaking-changes.md gains the b9621-b9637 rows.

https://claude.ai/code/session_01EQJCrQGmxCBf8WTCDuFE3X
@bernardladenthin bernardladenthin merged commit 20a38b5 into main Jun 14, 2026
8 of 10 checks passed
@bernardladenthin bernardladenthin deleted the claude/sweet-lamport-ugvqea branch June 14, 2026 20:49
@sonarqubecloud

Copy link
Copy Markdown

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.

2 participants