Added default labels '0' and '1'. Dealt with validation/test sets empty b/c too few samples.

Additions to README.me

Added ability to operate on a single node has no GPU; updated README accordingly.

Added text to README.md

Changed number of records process when testing from 100 to 10,000

Added --epoch option to bert_train_parallel.py

Some comments for launch.py

Partially updated README.md

Removed bert_train_political_leanings.py from repo

Added conditional on preponly where only sqlite existence was checked.

Moved label_encodings up above dataset creation..

Dataset creation even earlier.

ALlowed preponly without needing to set distributed sampling env vars.

Added --preponly to bert_train_parallel.py to just create the sqlite db from csv.

Switched result output files to easier formats: json and csv.