NumPy & SciPy for GPU

🔍 Analyze CUDA matrix multiplication performance and power consumption on NVIDIA Jetson Orin Nano across multiple implementations and settings.

High-performance CUDA implementation of Muon optimizer for LLM training. Features Newton-Schulz polar decomposition, cuBLAS acceleration, and transpose optimization for 8x FLOP savings on transformer FFN layers. Benchmarked on NVIDIA A100 with Llama 3.1 8B architectures (4096×11008 weights).

Rust製深層学習フレームワーク。Rustでゼロから実装して深層学習の原理を探究しよう！

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

Safe rust wrapper around CUDA toolkit

High-performance GPU-accelerated linear algebra library for scientific computing. Custom kernels outperform cuBLAS+cuSPARSE by 2.4x in iterative solvers. Built for circuit simulation workloads.

Cross platform C/C++ utilities.

Framework, toolkit and ready-to-use applications for numerical linear algebra dependent machine learning algorithms.

Scientific CUDA benchmarking framework: 4 implementations x 3 power modes x 5 matrix sizes on Jetson Orin Nano. 1,282 GFLOPS peak, 90% performance @ 88% power (25W mode), 99.5% accuracy validation, edge AI deployment guide.

Multiple GEMM operators are constructed with cutlass to support LLM inference.

🚀🚀🚀 This repository lists some awesome public CUDA, cuda-python, cuBLAS, cuDNN, CUTLASS, TensorRT, TensorRT-LLM, Triton, TVM, MLIR, PTX and High Performance Computing (HPC) projects.

SCUDA is a GPU over IP bridge allowing GPUs on remote machines to be attached to CPU-only machines.

gpumm - matrix-matrix multiplication by using CUDA, cublas, cublasxt and OpenACC.

Matrix multiplication on GPUs for matrices stored on a CPU. Similar to cublasXt, but ported to both NVIDIA and AMD GPUs.

VTensor, a C++ library, facilitates tensor manipulation on GPUs, emulating the python-numpy style for ease of use. It leverages RMM (RAPIDS Memory Manager) for efficient device memory management. It also supports xtensor for host memory operations.

SwiftCUBLAS is a wrapper for cuBLAS APIs with extra utilities for ease of usage, along with a suite of tests. The repository is tested on the newest (v12.5) CUDA runtime API on both Linux and Windows.

Library for profiling the execution time of CUDA official library functions

A Deep Learning framework with very few dependencies, Written in Rust

CUDA kernel functions