GitLab

Adds an example for loading a pre-trained BERT model and fine tune it as a language model (masked tokens & nextSentence) on your target corpus.

Minor README fix 

Fourth release

Change to use apex for better fp16 and multi-gpu support

fix compatibility with python 3.5.2

BertForMultipleChoice and Swag dataset example.

Merge branch 'fourth-release' of https://github.com/huggingface/pytorch-pretrained-BERT into fourth-release