CUDA 6大并行计算模式 代码与笔记

NVIDIA NVSHMEM is a parallel programming interface for NVIDIA GPUs based on OpenSHMEM. NVSHMEM can significantly reduce multi-process communication and coordination overheads by allowing programmer…

A fast communication-overlapping library for tensor/expert parallelism on GPUs.

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

Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.

This is a place for various problem detectors running on the Kubernetes nodes.

AIInfra（AI 基础设施）指AI系统从底层芯片等硬件，到上层软件栈支持AI大模型训练和推理。

gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI

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

DeepXTrace is a lightweight tool for precisely diagnosing slow ranks in DeepEP-based environments.

A lightweight parameter server interface

Tensors and Dynamic neural networks in Python with strong GPU acceleration

Nano vLLM

oneAPI Collective Communications Library (oneCCL)

主要记录大语言大模型（LLMs） 算法（应用）工程师相关的知识及面试题

Material for gpu-mode lectures

NCCL Tests

MSCCL++: A GPU-driven communication stack for scalable AI applications

Ongoing research training transformer models at scale

Example models using DeepSpeed

RDMA core userspace libraries and daemons

Optimized primitives for collective multi-GPU communication

DeepEP: an efficient expert-parallel communication library

2023年最新整理，qt开发最全面试集锦，含网络，文件系统，数据库，自定义控件，以及视频讲解，文档

Training/Evaluation code for HoVer-NeXt

游戏服务器框架，网络层分别用SocketAPI、Boost Asio、Libuv三种方式实现， 框架内使用共享内存，无锁队列，对象池，内存池来提高服务器性能。还包含一个不断完善的Unity 3D客户端，客户端含大量完整资源，坐骑，宠物，伙伴，装备, 这些均己实现上阵和穿戴, 并可进入副本战斗，多人玩法也己实现, 持续开发中。

k8s tutorials | k8s 教程

2023年最新整理 c++后端开发，1000篇优秀博文，含内存，网络，架构设计，高性能，数据结构，基础组件，中间件，分布式相关

系统设计面试：内幕指南（System Design Interview: An Insider’s Guide）