Comment EMA assert

[glenn-jocher]

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

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Refinement of Exponential Moving Average (EMA) code in YOLOv5's torch utilities.

📊 Key Changes

• Commented out an assertion check that ensured both EMA and model weights are in FP32 format.

🎯 Purpose & Impact

• This change potentially allows for more flexibility with data types used in model weights, facilitating experimentation and usage of different numerical precisions without triggering assertion errors.
• The impact on users could include fewer interruptions due to assertion failures, especially when working with mixed precision (FP16/FP32) models.