New Release: Ultralytics v8.2.94

Discussion
,
1

:star2: Ultralytics v8.2.94 Release Announcement

Hello Ultralytics Community!

We’re thrilled to announce the release of Ultralytics v8.2.94! This update brings a host of improvements and new features designed to enhance your experience and performance across various platforms. Here’s a quick overview:

:star2: Summary

This release enhances support for Apple’s MPS devices and includes various improvements for better performance and usability across platforms.

:bar_chart: Key Changes

  • Enhanced Memory Logging: Accurate GPU memory usage reporting for Apple’s MPS on macOS.
  • Improved Prediction Handling: Consistent bounding box handling and prediction saving.
  • Updated Documentation: Improved clarity and navigation for a better user experience.
  • Bug Fixes: Resolved issues with mps.empty_cache() on macOS without MPS.
  • New Benchmarks: Performance metrics for Intel’s latest hardware.

:dart: Purpose & Impact

  • :desktop_computer: macOS Improvements: Boosts training and inference performance on Apple hardware.
  • :chart_with_upwards_trend: Performance Benchmarks: Helps Intel users optimize their setups.
  • :page_facing_up: Documentation and Usability: Easier information access and contribution.
  • :triangular_ruler: Enhanced Prediction Handling: More robust and user-friendly model outputs.
  • :rocket: Optimized Platform Support: Ensures smoother operation across environments.

What’s Changed

  • Return boxes for SAM prompts inference by @Laughing-q
  • Docs improvements and redirect fixes by @glenn-jocher
  • Fix mps.empty_cache() for macOS without MPS by @Skillnoob
  • Add color palette tables to docs by @jk4e
  • Intel Core Ultra benchmarks by @ambitious-octopus
  • ultralytics 8.2.94 Apple MPS train memory display by @Oil3

Full Changelog: v8.2.93…v8.2.94

We encourage you to try out the new release and share your feedback. Your insights are invaluable to us and help drive future improvements.

Thank you for being a part of the Ultralytics community!

Release URL: Ultralytics v8.2.94

2

No matter what yolo model i use for train my macbook m3 pro is giving me this “NING :warning: NMS time limit 3.600s exceeded
Class Images Instances Box(P R mAP50 mWARNING :warning: NMS time limit 3.600s exceeded” failing to sort it I need to train large model with large dataset please help.

3

This is okay, there’s just a timeout for the NMS calculation so that it doesn’t get stuck. In an absolute ideal situation, it wouldn’t occur but it shouldn’t be a problem.

Of course, you can try a handful of basic troubleshooting steps:

  1. Restart your computer
  2. Make sure you’re using wall power when training (not battery)
  3. Ensure all other applications/processes are closed while training
  4. Update to the latest version of Ultralytics

Additionally, you could try using other versions of PyTorch than what you currently have installed or try using the Ultralytics Docker image for Arm64. Not that either of these would be certain to fix the issue, but they are worth trying if the basic steps above don’t help.

1 Like