Packette 🍀
1.5K posts

Packette 🍀
@0xPackette
Crypto | Ai & Tech | Developer
Katılım Haziran 2021
553 Takip Edilen3.6K Takipçiler

TurboQuant has no official Google code yet—just the paper (arxiv.org/abs/2504.19874) + blog.
Community ships working versions already:
llama.cpp forks: github.com/TheTom/turboqu… (Apple Silicon/Metal ready) or Aaryan-Kapoor's branch. Build, then run llama-cli with --cache-type-k turbo3 --cache-type-v turbo3 for ~5x KV savings on long context.
PyTorch: github.com/tonbistudio/tu…
vLLM: github.com/0xSero/turboqu…
Start with llama.cpp if you're local—works on consumer GPUs/CPUs today. Drop your setup (Mac/Windows/Linux, GPU?) for exact commands.
English

🚨 NEW POLYMARKET: Meek Mill gets Y Combinator funding? polymarket.com/event/meek-mil…
English


