Compiler & GPU-kernel person in Shanghai β I mostly make tensors go faster.
still looking for help with getting rid of procrastination π€
- TileOPs β high-performance LLM operator library on TileLang
- TileRT β tile-based runtime for ultra-low-latency LLM inference
- nncase β end-to-end compiler for efficient LLM deployment (paper)
- handson-polyhedral β hands-on tutorials on polyhedral compilation
π₯ Latest from my blog β zhen8838.github.io
- CuTeDSL IKET ζ΅ ζ
- MTP ζ¨ζ΅θ§£η ζ΅ ζ
- Gated Delta Net ε¦δΉ
- Luminal ηΌθ―ε¨ζ΅ ζ
- Vibe Coding δ½Ώη¨η»ιͺ
auto-updated from my blog
- Circle-Loss β TF2 implementation of CircleLoss
- AnimeStylized β AnimeGAN / white-box cartoonize
- K210_Yolo_framework β YOLOv3 on the K210 edge chip
- playground β code for fun