[Misc]feat: adapt to vLLM main (c7aa186d...2a16ece2)#239
Draft
Meihan-chen wants to merge 33 commits into
Draft
Conversation
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <jcccx.cmh@gmail.com>
Signed-off-by: Meihan-chen <239345480+Meihan-chen@users.noreply.github.com>
Signed-off-by: Meihan-chen <239345480+Meihan-chen@users.noreply.github.com>
Signed-off-by: Meihan-chen <239345480+Meihan-chen@users.noreply.github.com>
Signed-off-by: Meihan-chen <239345480+Meihan-chen@users.noreply.github.com>
Upstream range: c7aa186d67b6f051680831418e957c67f34ba7a2..16e336491e96041946e2cd379a80034fcac439bd Changes absorbed: - mistral_common version bump to 1.11.2 (vllm-ascend doesn't use directly) - Mistral tokenizer refactoring (removed adapt_inplace_to_mistral_tool) - Changed _prepare_apply_chat_template_tools_and_messages to _validate_apply_chat_template_args vllm-ascend adaptation: commit reference only (no code changes needed) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: main2main-bot <main2main-bot@users.noreply.github.com>
Upstream range: 16e336491e96041946e2cd379a80034fcac439bd..5d0fd87038b123a11dd9c85a05ce2d258e27ce7b Changes absorbed: - Build: libgomp fallback when torch has no vendored copy - CPU: FP8 W8A16 linear, dnnl AVX2 W8A8 Int8 - XPU: CUDA graph memory estimate fix, lora fixes - ROCm: gating tests, aiter attention, MoE fixes - Spec Decode: multimodal model warning instead of error - Bugfixes: PP sampled token receive, codegen, GLM4-MoE vllm-ascend adaptation: - Renamed _raise_if_multimodal to _warn_if_multimodal in dflash_proposer.py to match upstream llm_base_proposer.py change Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: main2main-bot <main2main-bot@users.noreply.github.com>
Upstream range: 5d0fd87038b123a11dd9c85a05ce2d258e27ce7b..242afc6bf40d6d088b7b97eda1dd7f00ddcdfa21 Changes absorbed: - Gemma4: Respect max_soft_tokens in encoder budget vllm-ascend adaptation: commit reference only (no Gemma4-specific code) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: main2main-bot <main2main-bot@users.noreply.github.com>
Upstream range: 242afc6bf40d6d088b7b97eda1dd7f00ddcdfa21..df8e63f4edabc06e51c10fe479fc9b95c303542f Changes absorbed: - nixl refactor: new transfer design - Removed MambaEngineTransferInfo, simplified register_remote_engine interface - kv_connector/utils.py refactoring vllm-ascend adaptation: commit reference only - vllm-ascend uses its own kv_transfer implementation (ascend_store, mooncake, etc.) - Does not use upstream TransferTopology or EngineTransferInfo interfaces Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: main2main-bot <main2main-bot@users.noreply.github.com>
Upstream range: df8e63f4edabc06e51c10fe479fc9b95c303542f..f39bcf1e30693f2e620e3a2ec732c5f7192408fb Changes absorbed: - OpenAI: tolerate empty content in forced tool choice - KV Offload: Return None from lookup() for in-flight blocks vllm-ascend adaptation: commit reference only - vllm-ascend uses upstream CPUOffloadingManager directly (inherits new lookup behavior) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: main2main-bot <main2main-bot@users.noreply.github.com>
Meihan-chen
force-pushed
the
main
branch
3 times, most recently
from
41aae44 to
9af7666
Compare
zhangxinyuehfad
force-pushed
the
main
branch
3 times, most recently
from
e4d6423 to
9f88e4b
Compare
MrZ20
force-pushed
the
main
branch
2 times, most recently
from
677396a to
9554149
Compare
zhangxinyuehfad
force-pushed
the
main
branch
5 times, most recently
from
c14c88e to
9768184
Compare
zhangxinyuehfad
force-pushed
the
main
branch
9 times, most recently
from
de24f70 to
60cd7c9
Compare
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
2 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.
Automated adaptation to upstream vLLM main branch changes.
Commit range: c7aa186d67b6f051680831418e957c67f34ba7a2...2a16ece2d342c0c154a4949ad317b521f8c04ec4
Created Commits
b5e2accc7f0a15c9059244ddd919690c57b3114bmain2main step-1: Update vLLM commit to 16e336491e59849ec341b3e47d6523ab6a9555fc8bd3e504main2main step-2: Update vLLM commit to 5d0fd8705a9a6e62094f44cea15932be88ba80dc5916b3d6main2main step-3: Update vLLM commit to 242afc6b455585a8ab37a782d3b4633ef36e5621277defe9main2main step-4: Update vLLM commit to df8e63f4db20b41a5c1366bbacfb3333fe46cffe56e5c541main2main step-5: Update vLLM commit to f39bcf1e3b47091e60c902d1cd557adaa75d801129054a16main2main step-6: Update vLLM commit to 27e0057a0a6df28f61b378f6313359077dd199e5f50bf95amain2main steps-7-14: Update vLLM commit to 2a16ece2main2main Summary
Main2Main Summary
Status: completed
Upstream range: c7aa186d67b6f051680831418e957c67f34ba7a2..2a16ece2d342c0c154a4949ad317b521f8c04ec4
Reached upstream commit: 2a16ece2d342c0c154a4949ad317b521f8c04ec4
Steps: 14/14
CI suite: e2e-main2main (not run in this environment)
Result
The target commit was fully reached. All 50 upstream commits across 14 planned steps were absorbed without code adaptation failures. The only vllm-ascend code change required was renaming
_raise_if_multimodalto_warn_if_multimodalinvllm_ascend/spec_decode/dflash_proposer.pyto match the upstream interface change inllm_base_proposer.py.Completed Steps
Changes Made
c7aa186d67b6f051680831418e957c67f34ba7a2to2a16ece2d342c0c154a4949ad317b521f8c04ec4._raise_if_multimodalto_warn_if_multimodalinvllm_ascend/spec_decode/dflash_proposer.py(step-2 adaptation).CI Verification
Key Adaptation Details
Step-2 - Speculative Decoding Interface Change:
_raise_if_multimodalto_warn_if_multimodalinvllm/v1/spec_decode/llm_base_proposer.pyAscendDflashProposerextendsAscendEagleProposerwhich extendsSpecDecodeBaseProposervllm_ascend/spec_decode/dflash_proposer.pyto maintain compatibilityNo adaptation needed for:
Follow-up
None. All upstream changes absorbed successfully. Recommend running full CI suite before merging to main.