Login / Signup

Integer Scale: A Free Lunch for Faster Fine-grained Quantization of LLMs.

Qingyuan LiRan MengYiduo LiBo ZhangYifan LuYerui SunLin MaYuchen Xie
Published in: CoRR (2024)
Keyphrases
  • fine grained
  • coarse grained
  • access control
  • tightly coupled
  • quantization error
  • scale space
  • databases
  • multiscale
  • massively parallel