* add pages about examples on training language models with fairseq and ray autoscaler
* better format
* update ray_train.sh
* Move EFS to the autoscaler file
* nits
* add comments to the code & use a new way to implement checkpoint hook
* small bug fix
* polish the doc
* fix formatting
* yaml
* update docs
* fix the bugs and add preprocess.sh
* fix lint
* Reduce batch size & fix lint
* shorttitle
