fix(deps): update dependency ultralytics ( 8.4.60 → 8.4.62 ) - #409

v8.4.62: - Prevent NaN/Inf EMA from discarding training checkpoints (#​24731)

Compare Source

🌟 Summary

🛡️ v8.4.62 is mainly a reliability release focused on preventing trained models from being lost at the end of training, with additional improvements to Platform docs, dataset/API documentation, test stability, and CI efficiency.

📊 Key Changes

• 🚨 Major training fix: checkpoints are no longer discarded just because EMA hits NaN/Inf during save checks

  • The most important change in this release, from PR #​24731 by @​glenn-jocher, fixes a bug where good models could finish training successfully but still fail to save any checkpoint.
  • This especially affected some runs using AdamW + AMP, where validation could corrupt the live EMA weights and cause repeated warnings like “Skipping checkpoint save... EMA contains NaN/Inf”.
  • The fix now:
    • keeps validation from modifying the live EMA in place
    • checks finiteness on the original fp32 EMA, not an already-converted fp16 copy
    • safely clamps overflow during checkpoint serialization instead of skipping the save

• ✅ Validation is now safer during AMP training

  • Validation still benefits from mixed precision speedups, but it no longer permanently “poisons” the EMA model.
  • This prevents a failure mode where one bad validation step could block checkpoint saving for the rest of training.

• 🧪 New test coverage for fp16 overflow checkpoint handling

  • Added tests to ensure models with large-but-finite EMA weights are still saved correctly.
  • This helps protect against regressions in future releases.

• 📘 Big Ultralytics Platform docs refresh

  • PR #​24726 by @​glenn-jocher significantly improved accuracy across Platform docs.
  • Updates include:
    • corrected UI labels and workflows
    • expanded Platform API reference
    • clearer dataset, annotation, deployment, training, billing, teams, and integrations docs
    • newly documented API capabilities like dataset embeddings, class management, GPU availability, import flows, and more

• 🔗 Fixed broken COCO evaluation links

  • PR #​24723 replaces dead COCO evaluation server links with the canonical COCO upload instructions.
  • Also fixes outdated docs author profile links.

• 🧪 Less flaky data-related tests

  • PR #​24724 reduces unnecessary downloads in tests and reuses cached assets when possible.
  • This should make CI more dependable and faster.

• ⚡ Lean CI improvements

  • PR #​24725 reduces git clone size and speeds up docs publishing and some test workflows.
  • PR #​24722 updates Codecov GitHub Actions from v6 to v7.

🎯 Purpose & Impact

• 💾 Prevents losing trained models

  • The headline fix is very important for users training YOLO models locally or in automated pipelines.
  • If your run trained well but ended with “no checkpoint was saved,” this release directly addresses that issue.

• 🔒 Improves training stability and trustworthiness

  • Users can have more confidence that successful training runs will actually produce saved checkpoints, especially when using AMP for faster training.

• 🚀 Better experience for common training setups

  • This is especially impactful for users training with AdamW + AMP, where the bug had been widely reported.
  • In practical terms: fewer surprise failures, less wasted compute, and less need for workarounds like disabling AMP.

• 📚 More accurate docs for the Ultralytics Platform

  • Platform users should now find the docs easier to follow and more aligned with what they actually see in the app.
  • This lowers confusion for both new and advanced users working with datasets, training, deployment, billing, and APIs.

• 🧰 Improved developer and CI reliability

  • Faster, lighter CI and more stable tests help maintain release quality and reduce false failures behind the scenes.

• 🌍 Cleaner external documentation links

  • Broken COCO benchmark links are fixed, making it easier for users to find the right evaluation submission path.

Overall, v8.4.62 is not a major model-feature release, but it is a high-value stability update 🛠️—especially for anyone training YOLO models with mixed precision and expecting reliable checkpoint saves.

What's Changed

• Fix broken COCO eval server and docs author links by @​glenn-jocher in #​24723
• Bump codecov/codecov-action from v6 to v7 in /.github/workflows by @​UltralyticsAssistant in #​24722
• Fix flaky data asset tests by @​glenn-jocher in #​24724
• Slim CI git clones: blobless docs publish, shallow Conda tests, drop unused checkout by @​glenn-jocher in #​24725
• Improve Platform docs accuracy and API coverage by @​glenn-jocher in #​24726
• Prevent NaN/Inf EMA from discarding training checkpoints by @​glenn-jocher in #​24731

Full Changelog: ultralytics/ultralytics@v8.4.61...v8.4.62

v8.4.61: - Fix INT8 ONNX calibration batch mismatch and read-only onnx2tf patching (#​24721)

Compare Source

🌟 Summary

v8.4.61 is mainly a stability and export reliability release 🛠️, led by an important fix for INT8 ONNX export failures and another fix for read-only onnx2tf patching, with additional improvements to CI, export testing, docs accuracy, and platform documentation.

📊 Key Changes

• 🚨 Fixed INT8 ONNX export crashes on small calibration datasets in PR #​24721 by @​glenn-jocher

  • Previously, exporting an INT8 ONNX model could fail if the calibration dataset had fewer images than the requested batch size.
  • This especially affected static-batch ONNX exports during quantization.
  • The fix now ensures calibration input batches match what the exported ONNX graph expects.

• 🔒 Fixed read-only onnx2tf patching issues in PR #​24721 by @​glenn-jocher

  • Another export bug was resolved for environments where files could not be patched because they were read-only.
  • This improves conversion reliability in stricter production or managed environments.

• 🧪 Stronger export validation in CI

  • PR #​24663 validates RKNN FP16 and INT8 export paths in CI using yolo26n.pt, helping catch export problems earlier.
  • Several CI workflow updates improve isolated export environments, dependency handling, and reduce flaky test failures on ARM runners.

• 🤖 TensorRT compatibility improvement

  • PR #​24647 automatically disables end2end export when using older TensorRT versions that do not support it.
  • This is especially useful for older Jetson setups and avoids hard-to-debug export/runtime failures.

• 🎯 SAM duplicate-mask cleanup fix

  • PR #​24688 fixes duplicate suppression in SAM post-processing when removing small regions.
  • This improves mask cleanup for affected SAM workflows.

• 🧠 Semantic segmentation support made clearer across the product

  • Platform docs now consistently describe semantic segmentation as a supported task.
  • Warning messages also now include task=semantic as a valid option.

• 📚 Large documentation and accuracy refresh

  • Many docs, examples, links, and model references were corrected.
  • YOLO26 docs were updated to align with the new paper and current behavior.
  • Platform docs were refreshed to reflect current features, export formats, integrations, and version requirements.

🎯 Purpose & Impact

• ✅ More reliable model export workflows

  • The biggest practical impact is that INT8 ONNX export should now work more reliably, even with small calibration datasets like quick test datasets.
  • This is especially important for users deploying optimized models to edge or production systems.

• 🚀 Fewer production export failures

  • The onnx2tf and ONNX calibration fixes target bugs already seen in real-world error tracking, so this release should reduce export breakages in actual deployments.

• 🧪 Better confidence in deployment formats

  • Expanded CI checks for RKNN, TensorRT, isolated export environments, and related dependencies mean export issues are more likely to be caught before release.

• 📦 Improved compatibility on specialized hardware

  • Users targeting Jetson, Rockchip, ARM systems, or other edge devices should see smoother behavior and fewer environment-specific failures.

• 🧹 Cleaner and more accurate user experience

  • SAM mask handling is improved in specific cases, warning messages are clearer, and docs now better match the actual product behavior.

• 🌍 Better guidance for a broad user base

  • Non-expert users benefit from clearer docs and more dependable defaults, while advanced users get more predictable export and deployment behavior across formats and devices.

In short, v8.4.61 is less about new models and more about making YOLO26 deployment safer, smoother, and more production-ready 🔧📈

What's Changed

• Validate RKNN FP16 export in CI by @​glenn-jocher in #​24663
• Fix tip case in docs by @​lakshanthad in #​24669
• Harden ARM CI against runner flakes by @​glenn-jocher in #​24665
• Sync Platform docs with current implement by @​laodouya in #​24671
• Update YOLO26 paper citations by @​glenn-jocher in #​24683
• fix: CLIP pkg_resources import failure on Python 3.8 by @​onuralpszr in #​24684
• Fix SAM remove_small_regions duplicate suppression by @​raimbekovm in #​24688
• Clarify YOLOE-26 visual prompt box format by @​glenn-jocher in #​24690
• Improve docs and docstring accuracy by @​glenn-jocher in #​24694
• Add 'semantic' to model task warning message by @​picsalex in #​24702
• Fix broken documentation links by @​glenn-jocher in #​24708
• Use watch URL for Tiger-Pose video links by @​glenn-jocher in #​24711
• Automatically disable end2end to be compatible with older versions of TensorRT<8.5.0 by @​yzfzzz in #​24647
• Fix broken link checker failures by @​glenn-jocher in #​24713
• Register GPU CI runner image at org level by @​glenn-jocher in #​24717
• Improve docs, comments, and correctness by @​glenn-jocher in #​24714
• Fix CLIP test asset cache location by @​glenn-jocher in #​24720
• Fix INT8 ONNX calibration batch mismatch and read-only onnx2tf patching by @​glenn-jocher in #​24721

New Contributors

• @​yzfzzz made their first contribution in #​24647

Full Changelog: ultralytics/ultralytics@v8.4.60...v8.4.61