Update train.py for val.run(half=amp)

[glenn-jocher]

Disable FP16 validation if AMP checks fail or amp=False.

May partially resolve #7908

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Improved validation inference with automatic mixed precision (AMP) during training.

📊 Key Changes

• Added the use of automatic mixed precision (AMP) to the validation step in the training process.

🎯 Purpose & Impact

• 🎯 Purpose: To enhance the efficiency and speed of the model validation phase by utilizing AMP.
• 💡 Impact: Users may experience faster validation times and reduced memory usage while training models, potentially improving the overall training performance on compatible hardware.