Load_encoder and load_pretrained don't return self

Following this topic : Why does learn.load return self?
After doing some NLP the other day, I noticed that both methods load_encoder and load_pretrained of the RNNLearner class perform their loading in place and return nothing. This is a different behavior than the ‘classic’ load method of a learner, I can do :

learn = text_classifier_learner(...).load(...)


learn = text_classifier_learner(...).load_encoder(...)

will return None.
This seems not optimal for the user to have such similar method behave differently. Is there a reason for those methods to return nothing, and if not, should they be changed to behave more like load?

It’s a bug, thanks for pointing it out. Will fix today!

Great ! I was wondering what was the best way to contribute to the library when finding things like that. Should I make a post on forum like this one that describes my finding, an issue on github, or even directly make a PR (that can be discarded is the issue appears not to be one) ?

Either is fine, though the PR is the fastest way to get it fixed (and like you said, if it’s not a real problem, we will just close the PR). Patch is ready, I’m just running the tests.