[Main2Main] Upgrade vllm commit to 0112#5691
[Main2Main] Upgrade vllm commit to 0112#5691zhangxinyuehfad wants to merge 1 commit intovllm-project:mainfrom
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request updates the vLLM commit dependency and adapts the codebase for the removal of init_cached_hf_modules in newer vLLM versions. The changes introduce a version check to maintain backward compatibility.
My review focuses on making this compatibility check more robust. I've suggested replacing the version-specific check with a feature detection approach (hasattr), which is a more resilient and maintainable pattern for handling optional functionalities across different library versions. I've also pointed out that the corresponding unit test should be made hermetic by mocking the version check, to ensure it's independent of the installed vLLM version.
github-actions
bot
added
documentation
|
👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:
If CI fails, you can run linting and testing checks locally according Contributing and Testing. |
zhangxinyuehfad
force-pushed
the
main0107
branch
from
40ed13a to
c861250
Compare
vllm-ascend-ci
added
ready
zhangxinyuehfad
force-pushed
the
main0107
branch
2 times, most recently
from
5224123 to
c046d9e
Compare
|
Another 4-cards break is due to: vllm-project/vllm#31773 |
zhangxinyuehfad
force-pushed
the
main0107
branch
from
c046d9e to
ef8e3f1
Compare
|
This pull request has conflicts, please resolve those before we can evaluate the pull request. |
zhangxinyuehfad
force-pushed
the
main0107
branch
from
ef8e3f1 to
62bc1f3
Compare
zhangxinyuehfad
force-pushed
the
main0107
branch
2 times, most recently
from
754b76d to
5f05e08
Compare
zhangxinyuehfad
changed the title
.github/workflows/_e2e_test.yaml
Outdated
| # TODO: add ignore after the issue is fixed | ||
| pytest -sv --durations=0 tests/e2e/singlecard/spec_decode/test_mtp_eagle_correctness.py | ||
| pytest -sv --durations=0 tests/e2e/singlecard/spec_decode/test_v1_spec_decode.py | ||
| pytest -sv --durations=0 tests/e2e/singlecard/spec_decode/test_v1_spec_decode.py \ |
There was a problem hiding this comment.
Please delete all skip cases, the break points are gone now
| from vllm.attention.backends.abstract import \ | ||
| AttentionMetadata # type: ignore | ||
| else: | ||
| from vllm.v1.attention.backend import AttentionMetadata # type: ignore |
There was a problem hiding this comment.
Please recheck all changes about this
zhangxinyuehfad
changed the title
zhangxinyuehfad
force-pushed
the
main0107
branch
from
caf4690 to
f2f9110
Compare
Signed-off-by: wjunLu <wjunlu217@gmail.com> Signed-off-by: hfadzxy <starmoon_zhang@163.com>
zhangxinyuehfad
force-pushed
the
main0107
branch
from
f2f9110 to
125d51b
Compare
|
main2main upgrade to 0113, The pr not need |
3 participants
What this PR does / why we need it?
Upgrade vllm commit to 0112 (d7b2e57097dae8a620c28eddf663adad2a8329c5)
init_cached_hf_modulesdue to [Chore] Try removeinit_cached_hf_modulesvllm#31786Does this PR introduce any user-facing change?
How was this patch tested?