Skip to content

feat: upgrade vllm amzn2023 to deepep v2 over efa - #6730

Draft
jinyan-li1 wants to merge 16 commits into
mainfrom
deepep-v2/vllm-amzn2023
Draft

feat: upgrade vllm amzn2023 to deepep v2 over efa#6730
jinyan-li1 wants to merge 16 commits into
mainfrom
deepep-v2/vllm-amzn2023

Conversation

@jinyan-li1

Copy link
Copy Markdown
Contributor

Upgrades the vLLM AL2023 EC2 + SageMaker images to DeepEP v2 over EFA.

Changes

  • vLLM 0.27.10.29.0 (NCCL 2.31.2 segfault fix)
  • DeepEP → AWS EFA fork amazon-contributing/DeepEP@874779c, built as a wheel from the fork (drops the upstream install_python_libraries.sh path)
  • NCCL 2.30.72.31.2 (pip wheel; required for the GIN/EFA backend)
  • NVSHMEM now via pip wheel nvidia-nvshmem-cu13==3.3.24
  • GDRCopy 2.5.2 userspace libs built from source (net-new for vLLM)
  • EFA 1.47.01.50.0 (auto --disable-ngc)
  • Add instanttensor + modelexpress
  • dlc_minor_version 4 → 5

Validation: DeepEP internode (test_ep.py on 2x p5en) pending capacity; gated test wiring to follow. Draft until validated.

@jinyan-li1
jinyan-li1 force-pushed the deepep-v2/vllm-amzn2023 branch from 059bf85 to 13b4dae Compare September 11, 2026 18:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant