good first issue
Metriche repository
- Star
- (1 stella)
- Metriche merge PR
- (Nessuna PR mergiata in 30 g)
Descrizione
Summary
Add a unit test to run the trainer with a very small config using a tiny model, ensuring that model loading, dataset preprocessing, and trainer setup succeed.
Acceptance Criteria
- A new test file
tests/test_trainer_small_config.pythat:- Uses
sshleifer/tiny-gpt2 - Builds a minimal JSON config
- Calls
CodeExplainerTrainer.load_model(),load_dataset(None),preprocess_dataset(), andsetup_trainer()
- Uses
- Test should run quickly (< 60s) and not perform full training.
References
- Existing tests in
tests/test_trainer.py - Tiny models used in CI