Update hubconf.py

[glenn-jocher]

This update prepares for PyTorch Hub page submission on https://github.com/pytorch/hub. PyTorch Hub tutorial has also been updated at #36

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Improvements in YOLOv5 model creation and checkpoint loading.

📊 Key Changes

• Introduced set_logging() at the beginning of hubconf.py to configure logging.
• Replaced string formatting for file paths with f-strings for better readability and performance.
• Streamlined the download and loading of pretrained model checkpoints.
• Added a condition to set model class names from the checkpoint if the number of classes matches.
• Included an example if __name__ == '__main__': block to demonstrate model creation.

🎯 Purpose & Impact

• 🧰 Enhanced code clarity and maintainability by using modern string formatting techniques.
• 💡 Simplified the process of loading pretrained weights, reducing potential for errors and improving user experience.
• 💼 Ensured model's class labels are accurately assigned when the checkpoint's class count aligns with the user's specified number of classes.
• 📚 Provided a handy example for users on how to instantiate a YOLOv5 model with pretrained weights.
• 🔄 Overall, these updates may lead to a more intuitive and error-free experience for developers working with YOLOv5 models.