Summary

Track all work required to build, validate, and publish CPU-enabled wheels for vLLM 0.16.x, and ensure they are integrated into the RHAIIS 3.4 EA2 delivery pipeline (indexes/images), with required runtime + import tests and hardware certification readiness.

Scope

In scope

• Select final upstream tag within vLLM 0.16.x line

• Build CPU wheels using approved CPU version for RHAIIS 3.4

• Ensure required CPU dev headers and toolchain present

• Publish wheels to AIPCC index

• Validate:

• • import tests
  • basic inference run
  • GPU detection
  • compatibility with targeted PyTorch

• Integrate into RHAIIS images/index for EA2

• Coordinate with certification + release engineering

Out of scope

• Major upstream feature backports beyond what's required for build/run

• Non-CPU accelerators (tracked separately)

Acceptance Criteria

• vLLM 0.16.x final target version agreed
• CPU wheels successfully built in AIPCC pipeline

• Wheels published to AIPCC index for RHAIIS 3.4 EA2

• Import test passes in RHAIIS container

• Basic inference test passes on supported

• Included in RHAIIS 3.4 EA2 compose

• Sign-off from CPU + RHAIIS release owners

Notes
This feature acts as the single tracking item for vLLM 0.16.x CPU enablement for RHAIIS 3.4 EA2.Sub-tasks should be created for build, dependency fixes, index publication, and validation.