Deprecate LightningModule.model_size

[roshikouhai]

What does this PR do?

This PR deprecates the model_size property in LightningModule as discussed and agreed in #8343

Does your PR introduce any breaking changes ? If yes, please list them.

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)
• Did you list all the breaking changes introduced by this pull request?

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 @roshikouhai! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

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

Comment last updated at 2021-07-30 13:19:42 UTC

[codecov]

Codecov Report

Merging #8495 (f8a6182) into master (5789e9f) will increase coverage by 0%.
The diff coverage is 100%.

@@          Coverage Diff           @@
##           master   #8495   +/-   ##
======================================
  Coverage      88%     88%           
======================================
  Files         168     168           
  Lines       13968   13975    +7     
======================================
+ Hits        12337   12344    +7     
  Misses       1631    1631           

[ananthsub]

Thanks for working on this @roshikouhai !

[tchaton]

LGTM !

[Borda]

if so, pls replace all usage in the codebase...

[roshikouhai]

Rebased on top of #8500. I tried updating test_quantization.py to use the new get_model_size() helper function but as @calebrob6 states, it currently doesn't work for models with sparse tensors which the tests have. Not sure if I should leave it be or find a way to implement the sparse tensors in this PR.

[calebrob6]

@roshikouhai, an alternate way to do get_model_size(...) is:

def get_model_size(self, model):
    with io.BytesIO() as f:
        torch.save(model.state_dict(), f)
        f.seek(0, io.SEEK_END)
        size_mb = f.tell() / 1e6
    return size_mb

I preferred to not do it this way as this will use quite a bit of RAM. Would you like me to look into how to implement getting the size of the sparse tensors as well?

[ananthsub]

@roshikouhai @calebrob6 - as the primary motivation for the PR is to remove this function off the core LightningModule API, what do you think about this sequencing?

1. Define the utility function which re-uses the exact same implementation and deprecate this off the core API and point users to the new utility
2. Later (e.g. for v1.5) investigate improvements we can make to implementation of the new utility to avoid file saving while supporting sparse tensors

[roshikouhai]

Later (e.g. for v1.5) investigate improvements we can make to implementation of the new utility to avoid file saving while supporting sparse tensors

Given the goal of this PR is to deprecate model_size from LightningModule, I vote towards this option and allow @calebrob6 the time he needs to implement the model_size for sparse tensors.

[ananthsub]

@Borda do you see any remaining issues? is this good to go?

[Borda]

fixed the remaining

[carmocca]

Pushed a commit to call the new function from the old one