deepspeedai / deepspeedai/DeepSpeed

[CUDA Kernel Profiling]

Open
#4,655 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement
Dominant language
Python
Stars
43.1k
Forks
5k
Avg merge
4d 15h
Merged PRs (30d)
112

Description

Thanks for all the great contributions to LLM training and inference.

I have a question regarding DeepSpeed team's development workflow for custom kernels in the context of the FastGen (and more broadly, as well). Specifically, how are kernels tested and profiled on the C++ / CUDA side before being bound to Python?

I see there are kernel-specific tests for ops and inference but these testing already built / bound kernels in Python. I'm interested in learning about tools / techniques and robust development practices for developing high-performance GPU kernels.

Any thoughts would be greatly appreciated!

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the kernel-specific tests under ops and inference and the FastGen custom-kernel workflow mentioned in the issue. Investigate how C++/CUDA kernels are tested and profiled before Python binding; done would be clear documentation of the team's tools and development practices.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp, python
Domain
documentation
Issue type
Documentation
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.