Skip to content

docker: upgrade dev vLLM image to 0.26.0 - #604

Draft
gc-fu wants to merge 2 commits into
mainfrom
codex/vllm-0.26-dev-image
Draft

docker: upgrade dev vLLM image to 0.26.0#604
gc-fu wants to merge 2 commits into
mainfrom
codex/vllm-0.26-dev-image

Conversation

@gc-fu

@gc-fu gc-fu commented Aug 7, 2026

Copy link
Copy Markdown
Contributor

Summary

Updates vllm/docker/Dockerfile.dev to build upstream vLLM v0.26.0 for Intel XPU.

  • Adds the Rust artifact build stage required by vLLM 0.26.
  • Installs upstream XPU requirements, including the Torch 2.12 dependency set and vllm_xpu_kernels 0.1.11.1.
  • Uses triton-xpu 3.7.1 and retains the offline oneCCL flow.
  • Removes the no-longer-needed multi-arc patches, custom ESIMD/private kernel builds, and BigDL-specific setup.

Compatibility

Keeps the existing intel/omix:0.1.0-devel-ubuntu24.04 base, proxy build arguments, /llm-scaler/vllm workspace, LD_LIBRARY_PATH, and offline oneCCL installation behavior.

Validation

  • git diff --check
  • Static Dockerfile consistency checks

A Docker build was not run because this checkout is itself inside a container without a Docker or BuildKit daemon.

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Updates the XPU development Docker image to build and install upstream vLLM v0.26.0 (including the new Rust frontend artifacts) while removing prior llm-scaler-specific multi-arc/kernel customization paths.

Changes:

  • Adds a separate rust-build stage to build vLLM Rust artifacts and copy them into the dev image.
  • Switches the dev image to build/install upstream vLLM v0.26.0 using upstream XPU requirements.
  • Updates dependency handling to enforce Triton-XPU 3.7.1 and retain the offline oneCCL installation flow.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment on lines +133 to +135
uv pip uninstall triton triton-xpu \
&& uv pip install triton-xpu==3.7.1 \
&& uv pip uninstall oneccl oneccl-devel \
Comment thread vllm/docker/Dockerfile.dev Outdated
# lean-core serving deps (same versions as prod) + bigdl-core (ships
# vllm_int4_for_multi_arc.so used by the sym_int4 path).
# ---------------------------------------------------------------------------
ENV LD_LIBRARY_PATH=/opt/intel/oneapi/ccl/2021.15/lib:/opt/intel/oneapi/mpi/2021.15/lib:/opt/intel/oneapi/compiler/2025.3/lib:/usr/local/lib
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants