pypi ultralytics 8.3.209
v8.3.209 - `ultralytics 8.3.209` RTDETR TFLite export fix with ONNX opset constraints (#22314)

latest releases: 8.3.211, 8.3.210
one day ago

🌟 Summary

RT-DETR exports to TensorFlow/TFLite are now reliable by automatically using ONNX opset 19, plus a small tuner bug fix and documentation/CI updates. βœ…πŸš€

πŸ“Š Key Changes

  • RT-DETR TensorFlow/TFLite export made robust

    • Automatically enforces opset=19 during the ONNX step when exporting RT-DETR to TensorFlow SavedModel (and downstream TFLite). See the PR by @Y-T-G: RTDETR TFLite export fix with ONNX opset constraints fixing issue #18055. πŸ”§
    • ONNX simplification remains enabled (simplify=True), improving export stability.
    • Scope is limited to the TensorFlow SavedModel export path; YOLO11/YOLO26 and other models/exports are unaffected.
  • Dependency updates for export

    • Bumped onnxslim to >=0.1.71 for both ONNX and TensorFlow SavedModel export paths for better model graph simplification. πŸ“¦
  • Tuner reliability fix

    • Correct best-iteration detection when resuming hyperparameter tuning, ensuring the β€œbest run” is identified accurately. Thanks @Y-T-G. 🧠
  • CI and Docs improvements

    • Temporarily disabled GPU CI jobs to avoid flaky runs due to runner availability. πŸ› οΈ
    • Added CI badges for more repos in CI Help Docs, including Ultralytics Stars and CLIP. βœ…
    • Updated docs banner to a more reliable PNG from the Ultralytics assets repo and added a β€œShenzhen” label with responsive styling for better branding. 🎨

🎯 Purpose & Impact

  • More reliable RT-DETR β†’ TensorFlow/TFLite exports

    • Prevents failures seen with higher ONNX opsets (e.g., 20+), improving out-of-the-box conversions. πŸ›‘οΈ
    • May override a user-specified higher opset to 19 for RT-DETR to ensure successful export. βš™οΈ
  • Smoother export experience

    • Newer onnxslim enhances ONNX graph simplification, reducing export friction. ⏩
  • More accurate tuning results

    • Fixes edge cases when resuming tuning so the best iteration is correctly tracked and saved. πŸ“ˆ
  • Clearer docs and steadier CI

    • Improved branding and asset reliability in docs, plus centralized CI visibility across more Ultralytics repos. πŸ“š
    • Temporary GPU CI pause improves pipeline stability but may reduce GPU regressions caught until re-enabled. ⚠️

Quick examples:

  • CLI: yolo export model=rtdetr.pt format=tf or yolo export model=rtdetr.pt format=tflite
  • Python:
    from ultralytics import YOLO
    
    model = YOLO("rtdetr.pt")
    model.export(format="tf")      # TensorFlow SavedModel
    model.export(format="tflite")  # TFLite

What's Changed

Full Changelog: v8.3.208...v8.3.209

Don't miss a new ultralytics release

NewReleases is sending notifications on new releases.