Announcing Ultralytics v8.3.9 Release!
Summary
We’re thrilled to announce the release of Ultralytics v8.3.9! This update brings significant improvements to training logs and CSV handling, enhancing your ability to track and report on model training sessions effectively.
Key Changes
- Version Update: From 8.3.8 to 8.3.9.
- Improved Training Logs: Enhanced logging for better tracking of epochs and training durations.
- CSV Enhancements: Training time is now included in CSV outputs for easier analysis.
- Batch Size Adjustment: Dynamic batch size adjustments for GPUs with over 16GB memory.
- Branding Updates: Default project and task identifiers renamed from “YOLOv8” to “Ultralytics”.
- Editable Installations: Removed
-e
pip install mode from CI workflows.
Purpose & Impact
- Enhanced Time Tracking: Detailed logging of training time helps manage and evaluate model training efficiency.
- Better Metric Tracking: Including training time in CSV outputs simplifies performance tracking and analysis.
- Optimized GPU Use: Larger batch sizes on high-memory GPUs enhance resource utilization for faster training.
- Consistent Branding: Unified naming promotes a cohesive brand identity across platforms.
- Improved Code Maintenance: Simplified installations and configurations enhance reliability.
What’s Changed
- Rename loggers from YOLOv8 to Ultralytics by @glenn-jocher in PR #16784
- AutoBatch larger batches on larger GPUs by @glenn-jocher in PR #16785
- Fix autobackend.py by @lmk123568 in PR #16801
- Update
speed-estimation
solution by @RizwanMunawar in PR #16798 - Remove -e tag from CI by @glenn-jocher in PR #16797
- Fix wrong parameter count while profiling by @Y-T-G in PR #16790
- Add training
time
to results CSV by @glenn-jocher in PR #16803
New Contributors
- @lmk123568 made their first contribution in PR #16801
Full Changelog: View here
We encourage you to try out the new release and share your feedback. Your insights are invaluable in helping us improve and evolve. Thank you for being a part of the Ultralytics community!