Updated Fabric trainer example to not call self.trainer.model
during validation
#19993
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Outdated suggestions cannot be applied.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.
Before, using the trainer would result in:
AttributeError: Your LightningModule code tried to access
self.trainer.model
but this attribute is not available when using Fabric with a LightningModule.Which would occur during validation when 'on_validation_model_eval' and 'on_validation_model_train' were called, as they do not work with fabric. This change allows the model to be switched between train and eval mode (which the original behavior intends) without relying on calling 'self.trainer.model'. A check is added for overridden methods to not change a user's code.
Fixes #19992
📚 Documentation preview 📚: https://pytorch-lightning--19993.org.readthedocs.build/en/19993/