cuda-kernels

Modern CUDA Learn Notes with PyTorch for Beginners, 200+ CUDA Kernels, Tensor Cores, HGEMM, FA-2 MMA.

Samples for CUDA Developers which demonstrates features in CUDA Toolkit

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

Ecosystem of libraries and tools for writing and executing fast GPU code fully in Rust.

LLM speculative inference server for heterogeneous hardware & consumer GPUs

CUDA Core Compute Libraries

Deep learning in Rust, with shape checked tensors and neural networks

Safe rust wrapper around CUDA toolkit

A machine learning library with a TypeScript API and Rust backend. CUDA and WebGPU compatibility. Built to understand how ML frameworks and models work internally.

CUDA Kernel Benchmarking Library

cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.

Pure Rust + CUDA LLM inference engine — no PyTorch, OpenAI-compatible, serves Qwen3 to Kimi-K2

FlashRT is a high-performance realtime inference engine for small-batch, latency-sensitive AI workloads. The flagship integration is production VLA control for Pi0, Pi0.5, GROOT N1.6, and Pi0-FAST. Also support llm e.g, qwen3.6-27B

CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning

Kernel Tuner

Simple utilities to enable code reuse and portability between CUDA C/C++ and standard C/C++.

Comprehensive CUDA tutorials for Maths & ML with examples

This is an archive of materials produced for an introductory class on CUDA programming at Stanford University in 2010

Amplifier allows .NET developers to easily run complex applications with intensive mathematical computation on Intel CPU/GPU, NVIDIA, AMD without writing any additional C kernel code. Write your function in .NET and Amplifier will take care of running it on your favorite hardware.

Triton implementation of FlashAttention2 that adds Custom Masks.