New
haochenglu/pi-attn — an inference engine targeting optimization and reproducibility
haochenglu/pi-attn is a Rust-based language model serving engine that fixes the order of operations so results stay reproducible even with batching and caching.
haochenglu/pi-attn
haochenglu/pi-attn is a Rust-based language model serving engine that fixes the order of operations so results stay reproducible even with batching and caching.
View on GitHub ↗- stars
- ★ 3
- Forks
- 0
- Language
- Rust
- License
- —
Posts
1No matching stories.