New
paddlepaddle/halops — LLM Operators for Blackwell SM100
paddlepaddle/halops is a library that provides attention and matrix multiplication operator paths for LLM inference, targeting NVIDIA Blackwell SM100.
paddlepaddle/halops
paddlepaddle/halops is a library that provides attention and matrix multiplication operator paths for LLM inference, targeting NVIDIA Blackwell SM100.
View on GitHub ↗- stars
- ★ 23
- Forks
- 8
- Language
- Cuda
- License
- Apache-2.0
Posts
1No matching stories.