New
surya-koritala/glyd — Lossless Compression of bf16 Weights and KV Cache
A look at the lossless GPU memory compression approach in surya-koritala/glyd and the measured results published in its README.
Projects
surya-koritala/glyd
Lossless AI compression: 33% less GPU memory, bit for bit.
View on GitHub ↗- stars
- ★ 252
- Forks
- 2
- Language
- Rust
- License
- NOASSERTION
- Latest release
- v0.32.0
Posts
1No matching stories.