Refactor SegformerFinetuner to Use on_validation_epoch_end Hook

Open SRDdev opened this issue 1 year ago • 1 comments

Description

This PR refactors the SegformerFinetuner class to ensure compatibility with PyTorch Lightning v2.0.0 by replacing the deprecated validation_epoch_end method with the on_validation_epoch_end hook. This change addresses the issue of maintaining compatibility with the latest PyTorch Lightning API.

Motivation:

PyTorch Lightning v2.0.0 has deprecated the validation_epoch_end method, necessitating this update.
This update ensures that validation metrics are computed and logged correctly using the new on_validation_epoch_end hook.

Relevant Links:

PyTorch Lightning v2.0.0 release notes

Dependencies:

None

Type of change

[x] Bug fix (non-breaking change which fixes an issue)
[ ] New feature (non-breaking change which adds functionality)
[ ] This change requires a documentation update

How has this change been tested, please provide a testcase or example of how you tested the change?

Unit Testing: Ensured that all existing unit tests for the SegformerFinetuner class pass with the new implementation.
Integration Testing: Conducted end-to-end tests to verify that the validation metrics are correctly computed and logged at the end of each epoch.
Validation Outputs: Confirmed that validation outputs are properly cleared after each epoch to prevent data leakage.

Any specific deployment considerations

No specific deployment considerations are required for this change. The updates are backward-compatible and do not introduce breaking changes.

Docs

[ ] Docs updated? What were the changes:
- No documentation updates were necessary as the changes are internal to the class implementation and do not affect the public API.