LLM training in simple, raw C/CUDA

Instant neural graphics primitives: lightning fast NeRF and more

Fast parallel CTC.

Fully Convolutional Instance-aware Semantic Segmentation

GPU database engine

Efficient GPU kernels for block-sparse matrix multiplication and convolution

Learn CUDA Programming, published by Packt

Examples demonstrating available options to program multiple GPUs in a single node or a cluster

Distributed multigrid linear solver library on GPU

A simple high performance CUDA GEMM implementation.

A CUDNN minimal deep learning training code sample using LeNet.

FLAME GPU 2 is a GPU accelerated agent based modelling framework for CUDA C++ and Python

PyTorch bindings for CUTLASS grouped GEMM.

webgpu GPU code implementation, including CUDA, OpenCL, OpenACC and C++ AMP