ultralytics 8.2.35

v8.2.35 - `ultralytics 8.2.35` add YOLOv9t/s/m models (#13504)

on Python PyPI

🌟 Summary

The v8.2.35 release of Ultralytics introduces new YOLOv9 models and various enhancements to the codebase and documentation.

📊 Key Changes

• New Model Configurations: Added configurations for YOLOv9 models (yolov9t, yolov9s, yolov9m).
• Documentation Updates: Enhanced model documentation with links to model weights and expanded on YOLOv9 details.
• Code Refactoring: Replaced Silence layers with nn.Identity in model configurations.
• New Modules: Introduced ELAN1 and AConv modules.
• Disabled CUDA Tests: Temporarily disabled CUDA export tests due to limited GPU availability.
• MLflow Sanitization: Added a sanitize_dict function for cleaning MLflow metrics logging.

🎯 Purpose & Impact

• Improved Model Availability: The new YOLOv9 configurations make it easier for users to access and utilize various model sizes tailored to different computational needs, enhancing flexibility and performance tuning.
• Enhanced Documentation: Links to model weights and better documentation provide users with more resources and clearer information, aiding in model selection and deployment.
• Code Quality: Refactoring and new modules ensure the codebase is more maintainable, robust, and clear, benefiting long-term development and user comprehension.
• Testing Adjustments: Disabling CUDA tests prevents potential issues due to resource constraints, ensuring smoother development and testing phases.
• Better Metrics Logging: The sanitize_dict function improves the accuracy and readability of logged metrics in MLflow, leading to better monitoring and analysis of model training.