Returning callback results when calling pipelines' train method

[VolodyaCO]

Closes #62.

This PR is a proof of concept for returning values from callbacks which might be useful for immediate manipulation after the pipeline has been run.

[clementchadebec]

Hi @VolodyaCO, sorry to come back to you so late.
Thank you very much for this contribution.

I was wondering though if we should not save the metrics in an easily readable file such as a csv instead of storing them in a dict and returning them at the end of the training as you suggest. The other TrainingCallbacks do not indeed return anything at the on_train_end stage and thus neither does trainer.train() in most cases. Therefore, I wonder if this will create confusion for users.

Let me know what you think of this :)

[VolodyaCO]

No worries @clementchadebec . I prefer not having side-effects such as saving things to a file. However, I see that to keep the changes consistent, one would want other TrainingCallbacks to return something explicitly (and at the moment they return None).

It looks as though another simple solution is not to return anything and not to save anything to a file. When using such a callback, the user would access the callback history and that's it. What do you think?

[clementchadebec]

Hi @VolodyaCO, thanks for the suggestion.
I agree with you that this seems to be the best solution to keep things consistent since the metrics would be available in the history attribute anyway.

[VolodyaCO]

@clementchadebec sorry for the delay. I just removed the returns!

[clementchadebec]

Hi @VolodyaCO,
No worries!
Thank you so much for making these changes. Everything looks fine :)