New Release: Ultralytics v8.3.148

:rocket: Ultralytics v8.3.148 Release Announcement

:glowing_star: Summary

We’re excited to announce Ultralytics v8.3.148! This release focuses on improved reliability for model export workflows and enhanced error handling, making it easier to troubleshoot and maintain your custom YOLO pipelines. No core workflow changes—just smoother, more robust performance for everyone!

Check out the Ultralytics v8.3.148 Release for all the details.


:new_button: New Features & Improvements

Updated Export Dependencies

  • ONNX & TensorFlow Export Reliability:
    We’ve updated the onnxslim dependency to v0.1.56 (contributed by @glenn-jocher). This brings greater compatibility and stabilization for ONNX and TensorFlow SavedModel exports, giving you a more consistent experience when moving your models between platforms.

Robust Error Handling

  • Smarter Troubleshooting:
    Improved error messaging around common pitfalls such as missing Python modules and outdated NumPy installations.
    • Handle numpy._core error while loading dataset cache (PR #20904 by @Y-T-G), ensuring that if your environment has an incompatible NumPy version, you’ll know exactly what to fix—no guesswork required!

:hammer_and_wrench: Improvements

  • Clearer User Feedback:
    Errors related to dataset cache loading and model weight imports are now easier to understand and resolve, especially when dependencies are missing.

  • Up-to-date Export Stack:
    ONNX and TensorFlow export functionality leverages the latest improvements from the underlying onnxslim library.


:lady_beetle: Bug Fixes

  • NumPy-related Cache Loading Errors:
    Users encountering unexpected tracebacks during dataset cache loading due to NumPy core changes will now see actionable guidance on how to resolve the problem.

:megaphone: How You Can Help

We encourage everyone to upgrade to v8.3.148, try out the updated export functionality, and let us know how it works for you!

Your testing, questions, and feedback are vital for helping us make YOLO models even better for the entire community. Please share your experiences, suggestions, or any issues right here in the discussion—or open a ticket on Ultralytics GitHub Issues.


:memo: Full Changelog

Check out the full changelog for all the details.


Thank you for being an essential part of the YOLO community and for helping advance state-of-the-art vision AI! If you have any questions or need assistance with the new release, just let us know—we’re here to help!