search

intel / auto-round

Advanced Quantization Algorithm for LLMs. This is official implementation of "Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs"
https://arxiv.org/abs/2309.05516
Apache License 2.0
200 stars 19 forks source link

refine the code and the speedup is notable #240

Closed wenhuach21 closed 1 week ago

wenhuach21 commented 1 week ago

on 125m, 30% speedup 7B models, 15%-20% speedup