smart_optimizer() improved reporting

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Refinement of optimizer logging and argument naming in YOLOv5 training code.

📊 Key Changes

• Removed the logging of scaled weight decay from train.py.
• Renamed weight_decay parameter to decay in the smart_optimizer function to maintain consistency.
• Updated logging in smart_optimizer to provide clearer information including learning rate and weight decay.

🎯 Purpose & Impact

• The changes streamline the codebase, making it cleaner by removing unnecessary logging information.
• The renaming of the parameter to decay clarifies its purpose and creates a more intuitive interface.
• Enhanced logging when creating the optimizer gives users clear visibility of the learning rate (lr) and decay values for different optimizer parameter groups, improving the transparency of the model training process.
• Overall, these adjustments lead to a more comprehensible and maintainable codebase with minimal direct impact on end users' daily work, but with improvements to the clarity and quality of information available during training.