Puzzles for learning Triton

A compact implementation of SGLang, designed to demystify the complexities of modern LLM serving systems.

Accelerating MoE with IO and Tile-aware Optimizations

MoE training for Me and You and maybe other people

🚀 Efficient implementations of state-of-the-art linear attention models

Helpful kernel tutorials and examples for tile-based GPU programming

NanoGPT (124M) in 3 minutes

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

Post-training with Tinker

An open-source AI agent that brings the power of Gemini directly into your terminal.

Mirage Persistent Kernel: Compiling LLMs into a MegaKernel

📚LeetCUDA: Modern CUDA Learn Notes with PyTorch for Beginners🐑, 200+ CUDA Kernels, Tensor Cores, HGEMM, FA-2 MMA.🎉

An Efficient and User-Friendly Scaling Library for Reinforcement Learning with Large Language Models

A course of learning LLM inference serving on Apple Silicon for systems engineers: build a tiny vLLM + Qwen.

Nano vLLM

Minimalistic large language model 3D-parallelism training

Universal memory layer for AI Agents

kernels, of the mega variety

My learning notes for ML SYS.

Scalable toolkit for efficient model reinforcement

Material for gpu-mode lectures

Curated collection of papers in MoE model inference

Large Language Model (LLM) Systems Paper List

s1: Simple test-time scaling

LMDeploy is a toolkit for compressing, deploying, and serving LLMs.

Perplexity GPU Kernels

🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"

A high-performance distributed file system designed to address the challenges of AI training and inference workloads.