YOLOv5 AutoCache Update

[glenn-jocher]

@kalenmike @AyushExel YOLOv5 AutoCache PR

Usage is:

python train.py  # no image caching
python train.py --cache   # defaults to 'auto'
python train.py --cache auto  # applies ram caching if sufficient RAM available (50% safety margin)
python train.py --cache ram  # force ram caching
python train.py --cache disk  # force disk caching

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancement of image caching and memory checks in the YOLOv5 training pipeline.

📊 Key Changes

• Clarified the --cache argument help description for image caching options.
• Introduced memory requirement checks to determine feasibility of caching images in RAM.
• Changed image caching messages to show cached memory in gigabytes (GB) for better readability.
• Removed requirement checks from notebook initialization to improve startup time and performance.

🎯 Purpose & Impact

• Code Clarity: The simpler help message for the --cache argument makes it more understandable to users.
• Performance Optimization: By checking if there is enough available RAM before caching images, the system prevents potential out-of-memory errors, leading to a smoother user experience.
• User Information: Displaying cached memory in GB provides a clear indication of how much system resource the process consumes.
• The removal of the requirement checks from the notebook initialization streamlines the startup process when using YOLOv5 in notebook environments like Jupyter.