TensorRT --dynamic fix

[glenn-jocher]

May resolve #9688

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

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Improved ONNX export compatibility with TensorRT versions 7 and 8.

📊 Key Changes

• Removed False in the call to export_onnx function, now passing dynamic directly.
• Fixed indentation and formatting in the dynamic flag handling block.

🎯 Purpose & Impact

• 🛠 The update ensures better compliance with different TensorRT versions when exporting YOLOv5 models to ONNX format.
• ⚙️ The removal of the hardcoded False argument allows for the dynamic flag to be passed properly, improving the dynamic batching capabilities.
• 📈 This change could lead to more efficient model deployment for users who utilize TensorRT for inference, potentially enhancing performance and resource utilization.