[Main2Main] Upgrade vllm commit to 0108#5727
[Main2Main] Upgrade vllm commit to 0108#5727zhangxinyuehfad wants to merge 2 commits intovllm-project:mainfrom
Conversation
There was a problem hiding this comment.
Code Review
This pull request upgrades the vLLM commit hash and introduces compatibility shims for changes in the new version. The changes correctly identify areas that need adaptation, such as the import path for PAD_SLOT_ID and the usage of init_cached_hf_modules.
However, there are some critical and high-severity issues. The compatibility logic relies on an exact version match (vllm_version_is), which is brittle and will likely fail for versions other than the one specified. This needs to be replaced with more robust version range checks. Additionally, the conditional import logic is duplicated across multiple files, which impacts maintainability. I've left specific comments with suggestions to address these points by refactoring the version checking utility and centralizing compatibility imports.
978aeed to
cef6be4
Compare
|
👋 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. |
9ee9933 to
7fbcbb0
Compare
|
This pull request has conflicts, please resolve those before we can evaluate the pull request. |
Signed-off-by: hfadzxy <starmoon_zhang@163.com>
Signed-off-by: hfadzxy <starmoon_zhang@163.com>
7fbcbb0 to
6111062
Compare
|
We don't need this PR anymore, please just keep 0112 |
What this PR does / why we need it?
Upgrade vllm commit to 0108 (eac3b96)
init_cached_hf_modulesdue to [Chore] Try removeinit_cached_hf_modulesvllm#31786vllm.v1.attention.backends.utilsduo to [Chore] Migrate V0 attention utils vllm#31891seq_lens_cpuandnum_computed_tokens_cpuCommonAttentionMetadata properties vllm#31773 ([Bugfix] Keep all tensors to be on the same device vllm#31958 will fix)Does this PR introduce any user-facing change?
How was this patch tested?