Add global YOLOv5_DATASETS_DIR

[glenn-jocher]

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

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Updated dataset directory configuration with environment variable support.

📊 Key Changes

• Removed the hardcoded 'datasets' directory path assignment.
• Added support for a custom 'YOLOv5_DATASETS_DIR' environment variable to specify the datasets directory.

🎯 Purpose & Impact

• Purpose: Allows users to define a custom directory for YOLOv5 datasets through an environment variable, providing greater flexibility in dataset management.
• Impact: Users can now easily switch datasets locations without modifying the code, which is particularly useful when working across different environments or with limited permission settings. This change enhances the configurability and usability of the YOLOv5 model training pipeline. 🛠️