Abnormal results of mPLUG-DocOwl2

[ZeldaM1]

Hi, thanks for sharing this excellent work.

I'm trying to run the Model Inference step provided in this repo (https://github.com/X-PLUG/mPLUG-DocOwl/tree/main/DocOwl2). However, the output seems to be abnormal and can not even be called a normal sentence (shown below).

Could you please provide any hints on possible reasons? For instance, environment requirements, GPU memory...

Thanks!