Top 3 Python cuda-kernel Projects

lmdeploy

3 2,324 9.8 Python

LMDeploy is a toolkit for compressing, deploying, and serving LLMs.

Project mention: AMD May Get Across the CUDA Moat | news.ycombinator.com | 2023-10-06

I wouldn’t say ROCm code is “slower”, per se, but in practice that’s how it presents. References:
https://github.com/InternLM/lmdeploy
https://github.com/vllm-project/vllm
https://github.com/OpenNMT/CTranslate2
You know what’s missing from all of these and many more like them? Support for ROCm. This is all before you get to the really wildly performant stuff like Triton Inference Server, FasterTransformer, TensorRT-LLM, etc.
ROCm is at the “get it to work stage” (see top comment, blog posts everywhere celebrating minor successes, etc). CUDA is at the “wring every last penny of performance out of this thing” stage.
In terms of hardware support, I think that one is obvious. The U in CUDA originally stood for unified. Look at the list of chips supported by Nvidia drivers and CUDA releases. Literally anything from at least the past 10 years that has Nvidia printed on the box will just run CUDA code.
One of my projects specifically targets Pascal up - when I thought even Pascal was a stretch. Cue my surprise when I got a report of someone casually firing it up on Maxwell when I was pretty certain there was no way it could work.
A Maxwell laptop chip. It also runs just as well on an H100.
THAT is hardware support.

kernel_tuner

4 243 9.1 Python

Kernel Tuner

Project mention: Ask HN: What apps have you created for your own use? | news.ycombinator.com | 2023-12-12

I've created Kernel Tuner (https://github.com/KernelTuner/kernel_tuner) as a small software development tool, because I was writing a lot of CUDA and OpenCL kernels at the time. I didn't want to manually figure out what best thread block dimensions and work division among threads were on every GPU over and over again.
The tool evolved quite a bit since the first versions. I'm also using it for testing GPU code, teaching, and it has become one of the main drivers behind a lot of the research that I do.

WorkOS

workos.com sponsored

The modern identity platform for B2B SaaS. The APIs are flexible and easy-to-use, supporting authentication, user identity, and complex enterprise features like SSO and SCIM provisioning.
halutmatmul

3 201 9.4 Python

Hashed Lookup Table based Matrix Multiplication (halutmatmul) - Stella Nera accelerator

Project mention: Show HN: Stella Nera – Maddness Hardware Accelerator | news.ycombinator.com | 2023-11-21

NOTE: The open source projects on this list are ordered by number of github stars. The number of mentions indicates repo mentiontions in the last 12 Months or since we started tracking (Dec 2020).