There is a growing field of study concerned with investigating the inner working of large-scale transformers like BERT (that some call "BERTology"). Some good examples of this field are:
- BERT Rediscovers the Classical NLP Pipeline by Ian Tenney, Dipanjan Das, Ellie Pavlick: https://arxiv.org/abs/1905.05950
- Are Sixteen Heads Really Better than One? by Paul Michel, Omer Levy, Graham Neubig: https://arxiv.org/abs/1905.10650
- What Does BERT Look At? An Analysis of BERT's Attention by Kevin Clark, Urvashi Khandelwal, Omer Levy, Christopher D. Manning: https://arxiv.org/abs/1906.04341
In order to help this new field develop, we have included a few additional features in the BERT/GPT/GPT-2 models to help people access the inner representations, mainly adapted from the great work of Paul Michel (https://arxiv.org/abs/1905.10650):
- accessing all the hidden-states of BERT/GPT/GPT-2,
- accessing all the attention weights for each head of BERT/GPT/GPT-2,
- retrieving heads output values and gradients to be able to compute head importance score and prune head as explained in https://arxiv.org/abs/1905.10650.
To help you understand and use these features, we have added a specific example script: [`bertology.py`](./examples/bertology.py) while extract information and prune a model pre-trained on MRPC.
## Notebooks
## Notebooks
We include [three Jupyter Notebooks](https://github.com/huggingface/pytorch-pretrained-BERT/tree/master/notebooks) that can be used to check that the predictions of the PyTorch model are identical to the predictions of the original TensorFlow model.
We include [three Jupyter Notebooks](https://github.com/huggingface/pytorch-pretrained-BERT/tree/master/notebooks) that can be used to check that the predictions of the PyTorch model are identical to the predictions of the original TensorFlow model.
