Hi everyone!
Sorry guys I introduce an evaluation bug in the previous version until v1.16.0, which makes the ATEPC evaluation relatively higher.
Now I fix this critical problem and improve the inference performance in non-Latin languages like Chinese, Japanese, Korean, etc.
And you can browse the [new demo](https://huggingface.co/spaces/yangheng/Multilingual-Aspect-Based-Sentiment-Analysis) for the non-Latin language inference change.
Cheers! Thanks for usingPyABSA.
Heng