🌟 Summary

The v8.2.86 release focuses on improving model export functionality, expanding operating system compatibility, and refining code quality across the Ultralytics project.

📊 Key Changes

• 🛠️ Model Export Enhancements: Updated logging for export failures, streamlined export logic, and improved error handling.
• 💻 Windows Compatibility: Added comprehensive testing for Windows and adjusted dependencies to address issues with PyTorch.
• 🎨 Code Modernization: Implemented modern Python practices like f-strings and argument-less super().
• 🔢 Improved Dataset Handling: Refined calibration and data loading processes to ensure consistency and reliability.

🎯 Purpose & Impact

• Enhanced Export Reliability: By increasing log visibility and removing unnecessary checks, users can diagnose issues faster, ensuring smoother model deployments.
• Widened OS Support: Including Windows in the CI testing matrix broadens the platform support, making the tool more versatile for developers on different systems.
• Cleaner Codebase: Modernized code not only boosts maintainability but also provides minor performance gains, making the project easier for new contributors to navigate.
• Consistency in Model Performance: Adjusting data loaders and calibration methods enhances the accuracy and repeatability of experiments and deployments.

These changes collectively aim to improve user experience, increase software reliability, and enhance performance stability. 🚀

What's Changed

• PyUpgrade 3.8 updates by @glenn-jocher in https://github.com/ultralytics/ultralytics/pull/15941
• Fixed OpenVINO int8 dynamic export and other minor changes by @ambitious-octopus in https://github.com/ultralytics/ultralytics/pull/14872
• ultralytics 8.2.86 Windows torch==2.4.0 patch by @glenn-jocher in https://github.com/ultralytics/ultralytics/pull/15942

Full Changelog: https://github.com/ultralytics/ultralytics/compare/v8.2.85...v8.2.86