# BERT-Small-L-4_H-512_A-8
This is a port of the [BERT-Small model](https://github.com/google-research/bert) to Pytorch. It uses 4 layers, a hidden size of 512 and 8 attention heads.