NumPy & SciPy for GPU

Safe rust wrapper around CUDA toolkit

An open collection of methodologies to help with successful training of large language models.

An open collection of implementation tips, tricks and resources for training large language models

Best practices & guides on how to write distributed pytorch training code

Distributed and decentralized training framework for PyTorch over graph

Efficient Distributed GPU Programming for Exascale, an SC/ISC Tutorial

Federated Learning Utilities and Tools for Experimentation

NCCL Fast Socket is a transport layer plugin to improve NCCL collective communication performance on Google Cloud.

A Julia wrapper for the NVIDIA Collective Communications Library.

Sample examples of how to call collective operation functions on multi-GPU environments. A simple example of using broadcast, reduce, allGather, reduceScatter and sendRecv operations.

Python Distributed Non Negative Matrix Factorization with custom clustering

N-Ways to Multi-GPU Programming

High performance NCCL plugin for Bagua.

NCCL Examples from Official NVIDIA NCCL Developer Guide.

use ncclSend ncclRecv realize ncclSendrecv ncclGather ncclScatter ncclAlltoall

Experiments with low level communication patterns that are useful for distributed training.

Blink+: Increase GPU group bandwidth by utilizing across tenant NVLink.

Hands-on Labs in Parallel Computing

Summary of call graphs and data structures of NVIDIA Collective Communication Library (NCCL)