Dynamic normalization layer selection

[glenn-jocher]

Based on actual available layers. Torch 1.7 compatible, resolves #7381

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Improved normalization layer detection in training script.

📊 Key Changes

• Refactored the way normalization layers are defined by programmatically identifying any nn.Module that contains 'Norm' in its name.

🎯 Purpose & Impact

• Simplifies Code Maintenance: Automates the inclusion of new normalization layers from PyTorch in the future without manual updates.
• Ensures Consistency: Reduces the risk of human error in maintaining the list of normalization layers.
• Potential to Improve Model Training: By being able to recognize and handle all types of normalization layers, there may be subtle improvements in model optimization.