Meituan-AutoML / MobileVLM

Strong and Open Vision Language Assistant for Mobile Devices
Apache License 2.0
890 stars 64 forks source link

mobilevlm_v2 with lora #30

Closed rangmiao closed 4 months ago

rangmiao commented 4 months ago

Dear author, I would like to inquire whether mobilevlm_v2 can be trained using the LoRa method for language models, and if so, how significant is the impact on accuracy?

sxu1997 commented 4 months ago

Sure, under setting of mobilevlm v2 with increasing data samples, using LoRA drops 1-2 points on average performance compared with full parameter fintuning.

er-muyue commented 4 months ago

Hi, we are closing this issue due to the inactivity. Hope your question has been resolved. If you have any further concerns, please feel free to re-open it or open a new issue. Thanks!