Uploaded image for project: 'AI Platform Core Components'
  1. AI Platform Core Components
  2. AIPCC-4565

vllm 0.10.1.1: add accelerator (for CUDA) packages to the builder

XMLWordPrintable

    • Icon: Epic Epic
    • Resolution: Done
    • Icon: Undefined Undefined
    • None
    • None
    • Accelerator Enablement
    • None
    • vllm 0.10.1: add accelerator packages to the builder
    • False
    • Hide

      None

      Show
      None
    • False
    • Done
    • AIPCC-4199 - vLLM 0.10.1.1 on CUDA wheels tracker
    • AIPCC-4199vLLM 0.10.1.1 on CUDA wheels tracker
    • 0% To Do, 0% In Progress, 100% Done

      Here are the accelerator packages that we need to add to support vllm 0.10.1:

      • bump flashinfer-python to 2.11 (don't carry out 2.8 patches, they are not needed)
      • investigate whether we need to stop building flash-attn and use the version from VLLM.

              emacchi@redhat.com Emilien Macchi
              emacchi@redhat.com Emilien Macchi
              Frank's Team
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated:
                Resolved: