Currently, it's another gateway based on Istio/Envoy. (TODO: give it a better description)

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

oneAPI Collective Communications Library (oneCCL)

DeepLearning Framework Performance Profiling Toolkit

mperf是一个面向移动/嵌入式平台的算子性能调优工具箱

The simplest, fastest repository for training/finetuning medium-sized GPTs.

A library to analyze PyTorch traces.

ROCm Communication Collectives Library (RCCL)

Alveo Collective Communication Library: MPI-like communication operations for Xilinx Alveo accelerators

NCCL Profiling Kit

Unified Communication X (mailing list - https://elist.ornl.gov/mailman/listinfo/ucx-group)

TACCL: Guiding Collective Algorithm Synthesis using Communication Sketches

Open-source observability for your LLM application, based on OpenTelemetry

Collective communications library with various primitives for multi-machine training.

awesome llm plaza: daily tracking all sorts of awesome topics of llm, e.g. llm for coding, robotics, reasoning, multimod etc.

Unified Collective Communication Library

A fast GPU memory copy library based on NVIDIA GPUDirect RDMA technology

Ongoing research training transformer models at scale

A PyTorch Native LLM Training Framework

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

选字验证码破解，试验过网易和极验，破解率99

Tools for monitoring NVIDIA GPUs on Linux

Infrastructure Programmer Development Kit (IPDK) is an open source, vendor agnostic framework of drivers and APIs for infrastructure offload and management that runs on a CPU, IPU, DPU or switch.

Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.

Grumpy is a Python to Go source code transcompiler and runtime.

Pretrain, finetune and deploy AI models on multiple GPUs, TPUs with zero code changes.

GPUDirect Async support for IB Verbs

High-performance, GPU-aware communication library