echarlaix/

bert-base-uncased-qqp-f87.8-d36-hybrid

This model was created using the nn_pruning python library: the linear layers contains 36% of the original weights.

The model contains 50% of the original weights overall (the embeddings account for a significant part of the model, and they are not pruned by this method).

This model was fine-tuned from the HuggingFace model checkpoint on task, and distilled from the model textattack/bert-base-uncased-QQP. This model is case-insensitive: it does not make a difference between english and English.

A side-effect of block pruning is that some of the attention heads are completely removed: 54 heads were removed on a total of 144 (37.5%).

Pytorch model file size: 377MB (original BERT: 420MB)