UltralyticsAssistant
released this
22 Oct 16:31
·
2 commits
to main
since this release
π Summary
The release of Ultralytics version 8.3.20 introduces improvements for efficiency and usability in both training logging and Docker image compatibility.
π Key Changes
- W&B Integration Fix: Adjusted the Weights & Biases logging to prevent errors when plots are deliberately disabled by the user.
- Docker Update: Updated the base Docker image to use a more recent version of PyTorch.
- Pretrained Model Documentation: Added examples for using pretrained YOLO models with Open Images Dataset V7.
π― Purpose & Impact
- Improvement in Efficiency: The W&B logging fix optimizes the training process by avoiding unnecessary plotting operations, thereby saving computational resources. π
- Enhanced Compatibility: By updating the Docker's PyTorch version, users benefit from potential performance boosts and better support for current CUDA features, facilitating more efficient processing. π
- Better Usability: The addition of code examples for pretrained models makes it easier for users to implement sophisticated AI functionality without a steep learning curve, boosting productivity and accessibility in AI projects. π§βπ»
What's Changed
- Dockerfile FROM pytorch/pytorch:2.5.0-cuda12.4-cudnn9-runtime by @glenn-jocher in #17094
- Add Open Images Dataset V7 pretrained model usage examples by @Y-T-G in #17090
ultralytics 8.3.20
W&Bplots=False
logging fix by @Anzhc in #17088
New Contributors
Full Changelog: v8.3.19...v8.3.20