CoderPat / structured-neural-summarization

A repository with the code for the paper with the same title
MIT License
74 stars 26 forks source link

Default parameters and tuning #19

Closed ioana-blue closed 5 years ago

ioana-blue commented 5 years ago

Some pieces of the code are not friendly to hyperparameter tuning. For example, there is no way to do no LR decay the way the code is written now:

i.e., if args.lr_decay_rate is not None

will never evaluate to true because lr_decay_rate has a default value. This line should probably be replace with something like:

if args.lr_decay_rate

which allows the lr_decay_rate to be set to 0 and that if portion doesn't execute.