Add on_fit_epoch_end callback

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements to training callbacks for COCO dataset evaluations and improved results logging.

📊 Key Changes

• Added an additional callback condition specifically for the COCO dataset during the training phase.
• Updated the on_train_end callback to include the results in its parameters.

🎯 Purpose & Impact

• 🎲 Tailored Callbacks for COCO Dataset: The update enables custom behavior during training when the model is using the COCO dataset, likely to process evaluation results specifically for this common benchmark.
• 📈 Enhanced Results Logging: By including results in the on_train_end callback, more detailed information will be logged at the end of training, providing users with better insights into model performance.
• Both changes should improve the user's insight and control over the training process, particularly when working with the well-known COCO dataset, potentially leading to more informed decision-making and streamlined workflows.