fixes typing in pytorch_lightning/callbacks/stochastic_weight_avg.py #13685
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
donlapark
requested review from
williamFalcon,
tchaton,
carmocca,
Borda,
kaushikb11,
SeanNaren,
awaelchli,
justusschock and
rohitgr7
as code owners
akihironitta
added
community
for more information, see https://pre-commit.ci
otaj
reviewed
Jul 18, 2022
…m:donlapark/lightning into fix_typing_callback_stochastic_weight_avg merge with prior changes
Borda
approved these changes
Jul 18, 2022
Codecov Report
@@ Coverage Diff @@
## master #13685 +/- ##
=========================================
- Coverage 86% 74% -12%
=========================================
Files 327 330 +3
Lines 25505 27641 +2136
=========================================
- Hits 21903 20475 -1428
- Misses 3602 7166 +3564 |
otaj
approved these changes
Jul 25, 2022
kaushikb11
approved these changes
Jul 26, 2022
carmocca
reviewed
Jul 26, 2022
Comment on lines
+248
to
+255
| module.running_mean, # type: ignore[arg-type] | ||
| device=pl_module.device, | ||
| dtype=module.running_mean.dtype, # type: ignore[union-attr] | ||
| ) | ||
| module.running_var = torch.ones_like( | ||
| module.running_var, device=pl_module.device, dtype=module.running_var.dtype | ||
| module.running_var, # type: ignore[arg-type] | ||
| device=pl_module.device, | ||
| dtype=module.running_var.dtype, # type: ignore[union-attr] |
There was a problem hiding this comment.
IMO these shouldn't have been ignored, but instead added assertions that running_{mean,var} are not None
| module.momentum = None | ||
| module.num_batches_tracked *= 0 | ||
| module.momentum = None # type: ignore[assignment] | ||
| module.num_batches_tracked *= 0 # type: ignore[assignment, operator] |
There was a problem hiding this comment.
Same comment here about asserting that it's not None
| anneal_epochs=self._annealing_epochs, | ||
| anneal_strategy=self._annealing_strategy, | ||
| last_epoch=trainer.max_epochs if self._annealing_strategy == "cos" else -1, | ||
| self._swa_scheduler: _LRScheduler = cast( |
There was a problem hiding this comment.
The type annotation is redundant when you cast
Suggested change
| self._swa_scheduler: _LRScheduler = cast( | |
| self._swa_scheduler = cast( |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Fixes typing in pytorch in
pytorch_lightning/callbacks/stochastic_weight_avg.pyfrom #13445.Note:
isinstance(..., Callable)throwing an arg-type error (isinstance(..., collections.abc.Callable) complains about "_SpecialForm" python/mypy#6864).self._swa_epoch_startis afloat, it is expected to be converted tointby a call toon_fit_start(). However,mypydoes not see this and throws in a return-value error.torch.optim.swa_utils.SWALRhas not been type-annotated yet, so the type of the argument ofswa_lris automatically set tofloat._BatchNorm.running_mean,_BatchNorm.running_varand_BatchNorm.num_batches_trackedhave typesOptional[Tensor]. These attributes are initialized to be zero tensors viaself.reset_parameters()butmypydoes not see this.Does your PR introduce any breaking changes? If yes, please list them.
Before submitting
PR review
Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the review guidelines. In short, see the following bullet-list:
Did you have fun?
Make sure you had fun coding 🙃