FP16xINT4 LLM inference kernel that can achieve near-ideal ~4x speedups up to medium batchsizes of 16-32 tokens.

Accelerating MoE with IO and Tile-aware Optimizations

Helpful kernel tutorials, examples and SKILLs for tile-based GPU programming

compiler learning resources collect.

NVIDIA cuTile learn

Distributed Compiler based on Triton for Parallel Systems

cuTile is a programming model for writing parallel kernels for NVIDIA GPUs

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

Nano vLLM

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

A Survey of Reinforcement Learning for Large Reasoning Models

Code repo for efficient quantized MoE inference with mixture of low-rank compensators

A Unix-like Operating System

Algorithms implementation in C++ and solutions of questions (both code and math proof) from “Introduction to Algorithms” (3e) (CLRS) in LaTeX.