Improve classification comments

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancement of documentation and examples for the Ultralytics YOLOv5 classification models.

📊 Key Changes

• Updated training checkpoint details, now including weight_decay parameter.
• Transitioned from showcasing the validation of the YOLOv5s-cls model to the YOLOv5m-cls model on the ImageNet-1k dataset.
• Predict section now suggests using the pretrained YOLOv5s-cls model to predict a single image (bus.jpg).
• Clarified the Export section by specifying the export of YOLOv5s-cls, ResNet, and EfficientNet models.
• Various documentation updates across the classify sub-module for clarity and consistency in instructions.

🎯 Purpose & Impact

• The added weight_decay parameter in documentation helps users understand model training configurations better.
• Highlighting the YOLOv5m-cls model provides insights into using a different category of pre-trained weights, possibly indicating a more robust or accurate model.
• Clarified examples make it easier for users to follow how to run individual tasks such as validating, predicting, and exporting models.
• The changes can lead to better user adoption and efficient usage of the YOLOv5 classification models, spreading knowledge on how to engage with the classification aspect of the Ultralytics ecosystem. 🚀