GPU Mode Releases 100+ Hands-On Lectures on CUDA and Kernel Optimization

blaizedsouza · x · 2026-08-21

The gpu-mode/lectures GitHub repo aggregates over 100 hands-on lectures on GPU programming. Topics cover CUDA, Triton, FlashAttention, vLLM, NCCL, SGLang, and kernel optimization, diving into memory architecture, Tensor Cores, distributed communication, and high-performance LLM inference. Speakers are from Stanford, Princeton, Meta, NVIDIA, AMD, Intel, and the PyTorch team.

Original post →

More from Infra

Infra channel →