kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity
main
TensorRT-LLMs/cpp/tensorrt_llm/deep_ep/README.md
Tailing Yuan 85b4a6808d
Refactor: move DeepEP from Docker images to wheel building (#5534) 
Signed-off-by: Tailing Yuan <yuantailing@gmail.com>
2025-07-07 22:57:03 +09:00

431 B
Raw Permalink Blame History

How to generate nvshmem_fast_build.patch?

  1. Build the project without applying the nvshmem_fast_build.patch.
  2. Link NVSHMEM to DeepEP with one NVSHMEM object file omitted.
  3. Repeat step 2 until no more object files can be omitted.
  4. Remove the unused files from NVSHMEM's CMakelists.txt, and save the differences as nvshmem_fast_build.patch.

The script strip_nvshmem_helper.py automatically performs steps 2 and 3.