Deprecate torch 1.6.0 compat _non_persistent_buffers_set

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced model compatibility and cleaned up outdated code in YOLOv5.

📊 Key Changes

• 🧹 Removed compatibility code for older PyTorch version (1.6.0) in Conv class.
• ✔️ Ensured nn.Upsample class is compatible with newer PyTorch version (1.11.0) by adding a check for the recompute_scale_factor attribute.

🎯 Purpose & Impact

• 🔍 The removal of outdated compatibility code makes the codebase cleaner and potentially reduces maintenance overhead.
• 🚀 Ensuring compatibility with newer versions of PyTorch prevents future deprecation issues, likely improving user experience and future-proofing the code.
• ✅ Users can have confidence in using YOLOv5 with the latest tech, maintaining performance and functionality.

[glenn-jocher]

Verified with

!git clone https://github.com/ultralytics/yolov5 -b glenn-jocher-patch-1 # clone
%cd yolov5
%pip install -r requirements.txt torch==1.7.0 torchvision==0.8.1  # install