Update vLLM version to v0.9.1

[CICD-at-OPEA]

Update vLLM version to v0.9.1

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues found.

Scanned Files

None

[louie-tsai]

this is because core binding requires privileged mode to run numa_migrate_pages call. this PR solved the issue: vllm-project/vllm#19241. two options here:

1. use privileged mode to run vLLM.
2. disable core binding in vLLM by setting VLLM_CPU_OMP_THREADS_BIND=all and TP/PP=1

[Copilot]

Pull Request Overview

This PR updates the vLLM version from v0.9.0.1 to v0.9.1 across multiple test scripts, docker-compose files, and environment configuration files.

• Updated vLLM version in test scripts for WorkflowExecAgent, VisualQnA, HybridRAG, DocSum, CodeTrans, CodeGen, ChatQnA, AudioQnA, etc.
• Added a new environment variable (VLLM_CPU_OMP_THREADS_BIND) in docker-compose files.
• Updated the build script environment variable in .github/env/_build_image.sh.

Reviewed Changes

Copilot reviewed 19 out of 19 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
WorkflowExecAgent/tests/2_start_vllm_service.sh	Updated vLLM version.
VisualQnA/tests/test_compose_on_xeon.sh	Updated vLLM version.
HybridRAG/tests/test_compose_on_gaudi.sh	Updated vLLM version.
DocSum/tests/test_compose_on_xeon.sh	Updated vLLM version.
CodeTrans/tests/test_compose_on_xeon.sh	Updated vLLM version.
CodeGen/tests/test_compose_on_xeon.sh	Updated vLLM version.
ChatQnA/tests/test_compose_without_rerank_on_xeon.sh	Updated vLLM version.
ChatQnA/tests/test_compose_qdrant_on_xeon.sh	Updated vLLM version.
ChatQnA/tests/test_compose_pinecone_on_xeon.sh	Updated vLLM version.
ChatQnA/tests/test_compose_on_xeon.sh	Updated vLLM version.
ChatQnA/tests/test_compose_milvus_on_xeon.sh	Updated vLLM version.
ChatQnA/tests/test_compose_mariadb_on_xeon.sh	Updated vLLM version.
ChatQnA/tests/test_compose_faqgen_tgi_on_xeon.sh	Updated vLLM version.
ChatQnA/tests/test_compose_faqgen_on_xeon.sh	Updated vLLM version.
AudioQnA/tests/test_compose_on_xeon.sh	Updated vLLM version.
AudioQnA/tests/test_compose_multilang_on_xeon.sh	Updated vLLM version.
AudioQnA/docker_compose/intel/cpu/xeon/compose_multilang.yaml	Added VLLM_CPU_OMP_THREADS_BIND variable.
AudioQnA/docker_compose/intel/cpu/xeon/compose.yaml	Added VLLM_CPU_OMP_THREADS_BIND variable.
.github/env/_build_image.sh	Updated vLLM version in the environment export.

Comments suppressed due to low confidence (3)

AudioQnA/docker_compose/intel/cpu/xeon/compose_multilang.yaml:47

• [nitpick] Consider adding a comment or updating related documentation to clarify the purpose of the new VLLM_CPU_OMP_THREADS_BIND variable, ensuring maintainability of the configuration.

VLLM_CPU_OMP_THREADS_BIND: all

AudioQnA/docker_compose/intel/cpu/xeon/compose.yaml:43

• [nitpick] Consider adding a comment or updating related documentation to clarify the purpose of adding VLLM_CPU_OMP_THREADS_BIND to the service configuration for easier future maintenance.

VLLM_CPU_OMP_THREADS_BIND: all

.github/env/_build_image.sh:5

• Ensure that VLLM_FORK_VER (currently set to v0.6.6.post1+Gaudi-1.20.0) remains compatible with the updated vLLM version v0.9.1 to prevent any integration issues.

export VLLM_VER=v0.9.1

[yinghu5]

@louie-tsai thank you very much for the solution. Both are workable.
but consider several related factors, we may upgrade the vllm docker image next release.