A lightweight design for computation-communication overlap.

DeepEP: an efficient expert-parallel communication library

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

ScheMoE

kernels, of the mega variety

slime is an LLM post-training framework for RL Scaling.

[NeurIPS 2025] ClusterFusion: Expanding Operator Fusion Scope for LLM Inference via Cluster-Level Collective Primitive

Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels

A list of awesome compiler projects and papers for tensor computation and deep learning.

Enhanced compiler frontend. Support Auto Compute + Auto Schedule + Auto Tensorize for tensor compilers.

Distributed Compiler based on Triton for Parallel Systems

KernelBench: Can LLMs Write GPU Kernels? - Benchmark with Torch -> CUDA (+ more DSLs)

Github Pages template based upon HTML and Markdown for personal, portfolio-based websites.

傻瓜式教程——如何使用Clash翻墙

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

⏰ Collaboratively track worldwide conference deadlines (Website, Python Cli, Wechat Applet) / If you find it useful, please star this project, thanks~

Fastest kernels written from scratch

flash attention tutorial written in python, triton, cuda, cutlass

GEMM by WMMA (tensor core)

Inference code for Llama models

We invite you to visit and follow our new repository at https://github.com/microsoft/TileFusion. TiledCUDA is a highly efficient kernel template library designed to elevate CUDA C’s level of abstra…

Some microbenchmark practices

collection of benchmarks to measure basic GPU capabilities