fix lm-head quant issue at disable_quanted_input

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

172 stars 20 forks source link

fix lm-head quant issue at disable_quanted_input #117

Closed wenhuach21 closed 3 months ago