An Open Source Machine Learning Framework for Everyone

LLM inference in C/C++

Truly independent web browser

Distribute and run LLMs with a single file.

Official inference framework for 1-bit LLMs

Tensor library for machine learning

CUDA Templates for Linear Algebra Subroutines

Fast and minimalist 3D viewer.

Mirage: Automatically Generating Fast GPU Kernels without Programming in Triton/CUDA

Demonstration of various hardware effects on CUDA GPUs.

校招、秋招、春招、实习好项目，带你从零动手实现支持LLama2/3和Qwen2.5的大模型推理框架。

TiledCUDA is a highly efficient kernel template library designed to elevate CUDA C’s level of abstraction for processing tiles.

Development repository for the Triton-Linalg conversion

Modern OpenGL binding for python

Ultra low overhead NVIDIA GPU telemetry plugin for telegraf with memory temperature readings.