Expected Accuracy for fine tuning bert-base-cased model with Yelp data

I am new to using Hugging Face API. I am following below tutorial on Hugging Face to fine tune bert-base model with Yelp data. I have followed the tutorial as-is with no changes on small dataset (1000 samples in training dataset) and used the Trainer provided by Hugging face. The best evaluation accuracy I get is 0.579. What is the expected accuracy? Has anyone else tried this tutorial? What accuracy did you get?