Remove ModelSummary validation from train loop on_trainer_init

[ananthsub]

What does this PR do?

• Start adding typehints to training_loop.py beginning with on_trainer_init for better static analysis w/ mypy or Pyre
• Remove duplication for weights_summary across TrainLoop and Trainer - the validation and property setting was happening in both places

Fixes #<issue_number>

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

[pep8speaks]

Hello @ananthsub! Thanks for updating this PR.

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2021-03-21 03:34:59 UTC

[ananthsub]

@edenafek @awaelchli @justusschock the model summary could be another (small) component we could split out from the core lightning module and offer as a standalone utility. the constructor for the summary object could take:

• an nn.Module
• precision
• device
• mode

connecting it back into the trainer here would be easy. it's almost there now with the ModelSummary object

[carmocca]

Related idea: #4541

[awaelchli]

Btw it's already a standalone utility. It can be used like this:

summary = LayerSummary(model, mode=...)
print(summary)

As you say it could also support nn.Module.

[ananthsub]

by standalone i mean no dependence on the lightning module. right now the precision and example inputs are looked up via attributes of model inside of model summary, but these could be passed in explicitly to the model summary constructor. it's not a huge deal, but something small for people to see how lightning builds reusable components

[carmocca]

Is this ready to review? If yes: please fill out the checklist in the original post, otherwise mark it as [WIP]/convert to draft

[ananthsub]

Is this ready to review? If yes: please fill out the checklist in the original post, otherwise mark it as [WIP]/convert to draft

yes this is ready for review

[tchaton]

LGTM !