How can I run mPLUG-Owl2 in two 48Gb GPU?

X-PLUG / mPLUG-Owl

mPLUG-Owl: The Powerful Multi-modal Large Language Model Family

https://www.modelscope.cn/studios/damo/mPLUG-Owl

MIT License

2.33k stars 176 forks source link

How can I run mPLUG-Owl2 in two 48Gb GPU? #243

Closed jhCOR closed 2 months ago

jhCOR commented 2 months ago

Thanks for the detailed explanation about the model. However, I'm facing an issue where I can't run the model using the provided script (finetune_lora.sh) due to OOM errors. What can I do to successfully finetune mPLUG-Owl2??

( I have 2 x 48Gb RAM GPU )