BERT-pytorch-for-learning

This code is written in accordance with "https://zh-v2.d2l.ai/chapter_natural-language-processing-pretraining/bert.html"

Training BERT:

python main.py

Fine tuning BERT:

(i.e. Natural Language Inference. We can use the features extracted by BERT to feed into the MLP, and through training and fine-tuning, determine whether the latter sentence can be inferred from the previous sentence. )

This is generally divided into three types: entailment, contradiction, neutral.

python fine_tuning.py

Test BERT:

You can try looking at BERT's output to see if the text pair used for the test is contextual.

python test_bert.py