Add weight decay to argparser

[glenn-jocher]

Signed-off-by: Glenn Jocher glenn.jocher@ultralytics.com

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Improved configurability of the weight decay parameter for training in the YOLOv5 classification module.

📊 Key Changes

• The optimizer now uses a customizable decay parameter from the command line options.
• Added a --decay command line argument to specify weight decay.

🎯 Purpose & Impact

• Purpose: To allow users greater flexibility when setting up the weight decay during model training, which can help in preventing overfitting and improving generalization.
• Impact: Users can now fine-tune the weight decay according to their dataset and training regimen, potentially leading to better model performance. 🚀