-
remove all absolute paths
-
get rid of all hardcoded parameters
-
write docstrings for all the files
-
make into an installed package
-
write some unit tests
-
include argparse in train_model and predict
-
upload a more recent model (github has trouble fitting the large ones)