Feature: save complete training configuration to disk after training

[JKurjenmiekka]

Description

Adds functionality to save the complete training configuration to disk after model training completes. The configuration is saved as a JSON file in the output directory and includes the full training config, model config, model config type (base or large), effective training parameters, class names, and number of classes after training.

This addition is motivated by having an easily accessible persistent record for reproducibility and documentation purposes. I also expect it to be useful for ensuring compatibility or debugging issues when running predictions on previously trained older models, e.g. for model comparison.

No new dependencies are required for this change.

Type of change

• New feature (non-breaking change which adds functionality)

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

I used an existing (and verifiably working) training pipeline after replacing the 1.1.0 rfdetr installation in .venv with my modified fork using uv pip install -e <path/to/modified/package>. I ran a short single epoch training to verify that, when training completes, a training_config.json file with the desired contents has been created in the output directory:

Any specific deployment considerations

Nope.

Docs

No Docs updated.

[CLAassistant]

All committers have signed the CLA.

[JKurjenmiekka]

I have read the CLA Document and I sign the CLA.