Update deps

Retrain model after update of transformers and other packages

0.3

Checkpoint release

0.2

Fix predictions when input is shorter than max length

0.1

First French model trained on 160M tokens from common crawl.