promote awq format for asym quantizaiton

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

245 stars 20 forks source link

promote awq format for asym quantizaiton #233

Closed wenhuach21 closed 1 month ago

wenhuach21 commented 2 months ago

The issue is there are some models like phi that awq has not been supported yet