[Main2Main] Upgrade vLLM to 0303#6944
Merged
wangxiyuan merged 11 commits intovllm-project:mainfrom Mar 6, 2026
Merged
Conversation
Contributor
|
Note Gemini is unable to generate a summary for this pull request due to the file types involved not being currently supported. |
Contributor
|
👋 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. |
Contributor
|
This pull request has conflicts, please resolve those before we can evaluate the pull request. |
MrZ20
requested review from
MengqingCao,
Yikun,
realliujiaxu,
wangxiyuan,
whx-sjtu and
zzzzwwjj
as code owners
33 tasks
MrZ20
force-pushed
the
main_0302
branch
7 times, most recently
from
b919e3c to
b097c45
Compare
wangxiyuan
approved these changes
Mar 5, 2026
Root causes: - CudagraphDispatcher.dispatch() disable_full replaced with valid_modes/invalid_modes (PR #34102) - compile_or_warm_up_model() now returns float compilation_time (PR #35503) - MMEncoderAttention forward methods added sequence_lengths param (PR #35564) - Removed auto-forcing of +rms_norm for sequence parallelism (PR #35410) Upstream commit range: 15d76f74e2fdb12a95ea00f0ca283acf6219a2b7..6290470843c131681e3e1318ae71070a34f33225 Co-Authored-By: Claude Code <noreply@anthropic.com> Signed-off-by: MrZ20 <2609716663@qq.com>
Root causes: - CudagraphDispatcher.dispatch() API changed: disable_full -> valid_modes/invalid_modes (#34102) - compile_or_warm_up_model() must return float compilation_time (#35503) - MMEncoderAttention.forward_oot() gained new sequence_lengths param (#34580) - +rms_norm no longer auto-forced for SP, breaks Ascend without CUDA _C ops (#35410) Upstream commit range: 15d76f74e2fdb12a95ea00f0ca283acf6219a2b7..6290470843c131681e3e1318ae71070a34f33225 Co-Authored-By: Claude Code <noreply@anthropic.com> Signed-off-by: MrZ20 <2609716663@qq.com>
Signed-off-by: MrZ20 <2609716663@qq.com>
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
LCAIZJ
pushed a commit
to LCAIZJ/vllm-ascend
that referenced
this pull request
Mar 7, 2026
### What this PR does / why we need it? break: - vllm-project/vllm#34102 Disable_full param replaced with valid_modes/invalid_modes API - vllm-project/vllm#35503 Now must return float compilation_time - vllm-project/vllm#35564 New sequence_lengths param added - vllm-project/vllm#33807 A check was performed (if runner_backend != "auto") - vllm-project/vllm#34861 `BaseDeviceCommunicator` now accesses PyTorch's internal `pg_map` to check process group state - vllm-project/vllm#35274 **Important change:** - vllm-project/vllm#28672 `matcher_utils` directly accesses `torch.ops._C.*` during the import phase. In the Ascend environment, some unregistered ops trigger `AttributeError`, causing e2e initialization failure. https://github.com/vllm-project/vllm-ascend/actions/runs/22607260487/job/65502047131#step:10:2323 https://github.com/vllm-project/vllm/blob/main/vllm/compilation/passes/fusion/matcher_utils.py#L29 This PR adds temporary compatibility placeholders (rms_norm, fused_add_rms_norm, rotate_embedding, static/dynamic fp8 quant, silu_and_mul) to `vllm_ascend/patch/platform/patch_fusion_matcher_compat_ops.py` to ensure no crashes during the import phase. Upstream repairs will be considered later. ### Does this PR introduce _any_ user-facing change? ### How was this patch tested? - vLLM version: v0.16.0 - vLLM main: vllm-project/vllm@15d76f7 --------- Signed-off-by: MrZ20 <2609716663@qq.com> Signed-off-by: gcanlin <canlinguosdu@gmail.com> Co-authored-by: Meihan-chen <jcccx.cmh@gmail.com> Co-authored-by: Claude Code <noreply@anthropic.com> Co-authored-by: gcanlin <canlinguosdu@gmail.com>
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
5 participants
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.
What this PR does / why we need it?
break:
Disable_full param replaced with valid_modes/invalid_modes API
Now must return float compilation_time
New sequence_lengths param added
--moe-backendarg for explicit kernel selection vllm#33807A check was performed (if runner_backend != "auto")
BaseDeviceCommunicatornow accesses PyTorch's internalpg_mapto check process group statebc-lintvllm#35274Important change:
matcher_utilsdirectly accessestorch.ops._C.*during the import phase. In the Ascend environment, some unregistered ops triggerAttributeError, causing e2e initialization failure.https://github.com/vllm-project/vllm-ascend/actions/runs/22607260487/job/65502047131#step:10:2323
https://github.com/vllm-project/vllm/blob/main/vllm/compilation/passes/fusion/matcher_utils.py#L29
This PR adds temporary compatibility placeholders (rms_norm, fused_add_rms_norm, rotate_embedding, static/dynamic fp8 quant, silu_and_mul) to
vllm_ascend/patch/platform/patch_fusion_matcher_compat_ops.pyto ensure no crashes during the import phase. Upstream repairs will be considered later.Does this PR introduce any user-facing change?
How was this patch tested?