rename Model steps

[williamFalcon]

What does this PR do?

Fixes # (issue).

PR review

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.

Did you have fun?

Make sure you had fun coding 🙃

[pep8speaks]

Hello @williamFalcon! Thanks for updating this PR.

• In the file pytorch_lightning/core/lightning.py:

Line 453:101: E501 line too long (113 > 100 characters)
Line 465:101: E501 line too long (109 > 100 characters)
Line 517:101: E501 line too long (105 > 100 characters)
Line 530:101: E501 line too long (105 > 100 characters)
Line 533:101: E501 line too long (104 > 100 characters)
Line 580:101: E501 line too long (102 > 100 characters)
Line 590:101: E501 line too long (111 > 100 characters)
Line 591:101: E501 line too long (117 > 100 characters)
Line 677:101: E501 line too long (107 > 100 characters)
Line 689:101: E501 line too long (109 > 100 characters)

Comment last updated at 2020-03-05 17:32:23 UTC

[ethanwharris]

Minor comment in the docs, everything else looks great, and should be backwards compatible :)

[ethanwharris]

Perhaps these should say validation_epoch_end or validation_step_end?

[awaelchli]

Will validation_end be renamed to validation_epoch_end? Validation can be done multiple times per epoch by setting the flag in the trainer, so renaming that would be confusing.

[williamFalcon]

Will validation_end be renamed to validation_epoch_end? Validation can be done multiple times per epoch by setting the flag in the trainer, so renaming that would be confusing.

yes, but validation_epoch_end is called at the end of a validation epoch NOT training epoch.
But I could see how it might be confusing. Is it better to explain in docs?

the general new pattern is:

___step
___step_end
___epoch_end

__step_end is only necessary if you use dp and need full batch loss or something (ie: negative sampling)