Rebase Arctic Inference onto vLLM v0.26.0 - #269
Open
sfc-gh-mhidayetoglu wants to merge 12 commits into
Open
Conversation
sfc-gh-sbekman
requested changes
Aug 20, 2026
Co-authored-by: Stas Bekman <stas.bekman@snowflake.com>
Co-authored-by: Stas Bekman <stas.bekman@snowflake.com>
sfc-gh-sbekman
self-requested a review
August 24, 2026 23:07
sfc-gh-jrasley
approved these changes
Aug 25, 2026
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
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
pyproject.tomland adjust plugin/patches/model runner code for the newer vLLM APIs.ARCTIC_INFERENCE_ENABLED=1, a vLLM version mismatch now raises instead of silently running unpatched (ARCTIC_INFERENCE_SKIP_VERSION_CHECK=1bypasses).max_tokens_n) feature into the plugin patching system and drop the standalone source-patch scripts.skills/.Changes
plugin.py,patches.py,model_runner.py,args.py,ulysses.py, attention forest cascade, spec-dec fp8, SwiftKV llama.plugin.pyhard-fails on enabled+mismatch;utils.pyintroducesVLLM_PATCH_VERSIONas the single source of truth (decoupled from install extras);server/worker.pyengine-args selection simplified accordingly.arctic_inference/vllm/sampling.py(ParentRequestPatch) applied inapply_arctic_patches(); per-child lengths passed viaSamplingParams.extra_args["max_tokens_n"]. Removesbenchmark/rollout/{parallel_sampling.patch,sampling_params.patch,patch_sampling.sh}; updatesbenchmark/rollout/README.md.pyproject.toml,csrc/custom_ops/CMakeLists.txt.skills/arctic-inference-architecture,skills/arctic-vllm-versioning,skills/rebase-arctic-inference(+ example rebase plans/reports and helper scripts).tests/unit_tests/test_custom_ops.py.