Revert "ci: add runs-on/cache to skip C++ rebuild on PR CI (#469)"#481
Open
JavaPythonAIForBAT wants to merge 1 commit into
Open
Revert "ci: add runs-on/cache to skip C++ rebuild on PR CI (#469)"#481JavaPythonAIForBAT wants to merge 1 commit into
JavaPythonAIForBAT wants to merge 1 commit into
Conversation
…ct#469)" This reverts commit f3f7d23 (sgl-project#469). Removes the runs-on/cache-based wheel reuse from pr-test-npu.yml and deletes push_build_cache.yml. The cache scheme has two issues that together broke test-build-deepep-a2: 1. push_build_cache.yml's a2 job ran a bare `bash build.sh`, whose defaults produce an A3 wheel (Ascend910_9382 + csrc/deepep/ops), and stored it under the A2 cache key — so PR runs on A2 hardware restored an A3 wheel. 2. cache key strings hard-code cann8.5.0 while pr-test-npu.yml containers were upgraded to cann 9.0.0 in sgl-project#476, so even a correctly-built wheel would be loaded inside a mismatched runtime. Reverting now to unblock PR CI; the cache mechanism can be reintroduced once both issues are addressed. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Contributor
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
4 tasks
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Revert #469 to unblock PR CI. The cache scheme it introduced has two issues that together break
test-build-deepep-a2:push_build_cache.yml'sbuild-cache-a2runs a barebash build.sh, whose defaults areBUILD_DEEPEP_OPS=ONandSOC_VERSION=Ascend910_9382— i.e. an A3 wheel — and stores it under the A2 cache key. PR runs on A2 hardware then restore an A3 wheel and fail at test time.cann8.5.0, butpr-test-npu.ymlcontainers were upgraded tocann 9.0.0in cann9.0.0 Adapt #476. Even a correctly-built wheel would be loaded inside a mismatched runtime.This PR removes both:
push_build_cache.yml(deleted)Get build hash/Restore build cachesteps and the cache-hit branch inPrepare Deepepfor the threepr-test-npu.ymljobsContainer images / other CI behavior are preserved as-is.
The cache mechanism can be reintroduced in a follow-up once the wheel-architecture and CANN-version coupling are fixed (see also #480 for the
-a deepep2and key-bump portion).Test plan
test-all-build,test-build-deepep-a3,test-build-deepep-a2all run a fullbash build.sh ...(no cache) and reach the test stagesinternode/daily-build-test/build_and_releaseworkflows untouched🤖 Generated with Claude Code