Difference of performance when finetuning bert use the huggingface or the google official code

[Gordon]

Hi all,
I tend to finetune BERT with a simple text classification task.
However, I got different results when using the huggingface library (torch 1.8.1+cu111) and google’s official code (Tensorflow v1.15).
I wonder if there is any optimization in huggingface for fine-tuning bert?

By the way, I believe that I use the same hyper-parameters. But I got the higher performance using the huggingface library.

I have checked the details in a similar issue:

Is Google's official BERT model and huggingface BERT model different or same?

Google’s Official BERT HuggingFace BERT If they are same/equivalent, how are they created? Are the pretrained parameters just copied? Or are they pretrained from scratch following exact same approach (objective/hyperparameters) and data mention in the paper? I have been searching for this information for hours, but can’t find anywhere.

However, it does not help me.