Add DNN warning comment

[glenn-jocher]

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

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Updated ONNX export settings for compatibility with PyTorch versions.

📊 Key Changes

• Added a comment warning regarding the do_constant_folding parameter in ONNX export.
• No actual code behavior change—only additional information for developers.

🎯 Purpose & Impact

• 🎓 The comment is aimed at informing developers that using PyTorch version 1.12 or newer might require setting do_constant_folding=False during ONNX export for successful DNN inference.
• 🔧 This update is a proactive measure to prevent potential issues with ONNX model deployment, ensuring smoother transitions for users upgrading their PyTorch version.
• 🤖 Users are unlikely to see immediate changes, but the comment provides crucial insights for troubleshooting and future-proofing their ONNX model exports.