Skip to content

refactor(cuda & rocm): bump vllm to v0.20.2#157

Open
gitlawr wants to merge 2 commits into
gpustack:mainfrom
gitlawr:vllm20.2
Open

refactor(cuda & rocm): bump vllm to v0.20.2#157
gitlawr wants to merge 2 commits into
gpustack:mainfrom
gitlawr:vllm20.2

Conversation

@gitlawr
Copy link
Copy Markdown

@gitlawr gitlawr commented May 21, 2026

No description provided.

gitlawr added 2 commits May 21, 2026 15:19
Signed-off-by: gitlawr <lawrleegle@gmail.com>
Match the workflow behavior in .github/workflows/pack.yml: when
pack/<backend>/Dockerfile.<service> exists, use it instead of the
backend's main Dockerfile.

Signed-off-by: gitlawr <lawrleegle@gmail.com>
Copilot AI review requested due to automatic review settings May 21, 2026 09:24
Copy link
Copy Markdown

@gemini-code-assist gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

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

Code Review

This pull request updates the vLLM version from 0.20.1 to 0.20.2 across the project's Dockerfiles and build matrix configuration. It also modifies the Makefile to use the JOB_DOCKERFILE variable for specifying Dockerfile paths, allowing for more flexible build configurations. I have no feedback to provide.

Copy link
Copy Markdown

Copilot AI left a comment

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 container packaging configuration to build vLLM images against vLLM v0.20.2 for both AMD ROCm and NVIDIA CUDA, aligning default Dockerfile args and the build matrix; also fixes the packaging build to honor service-specific Dockerfiles.

Changes:

  • Bump vLLM base image tags and VLLM_VERSION from 0.20.1 to 0.20.2 for ROCm and CUDA packaging.
  • Update pack/matrix.yaml build args to use the v0.20.2 images across ROCm and CUDA variants.
  • Make make package use the computed service-specific Dockerfile path (Dockerfile.<service>) when present.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File Description
pack/rocm/Dockerfile.vllm Bumps default vLLM base image/version args to v0.20.2 for ROCm builds.
pack/cuda/Dockerfile.vllm Bumps default vLLM base image/version args to v0.20.2 for CUDA builds.
pack/matrix.yaml Updates matrix build args to vLLM 0.20.2 and corresponding base image tags.
Makefile Uses JOB_DOCKERFILE so packaging builds can select Dockerfile.<service> when available.

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

Comment thread pack/matrix.yaml
Comment on lines 60 to +64
- "linux/amd64"
args:
- "ROCM_VERSION=7.2.1"
- "VLLM_VERSION=0.20.1"
- "VLLM_BASE_IMAGE=vllm/vllm-openai-rocm:v0.20.1"
- "VLLM_VERSION=0.20.2"
- "VLLM_BASE_IMAGE=vllm/vllm-openai-rocm:v0.20.2"
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