ultralytics 8.1.4 RTDETR TensorBoard graph visualization fix

[glenn-jocher]

May resolve #7714

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements to the Ultralytics AI software for improved documentation, dataset handling, model stability, and Tensorboard integration.

📊 Key Changes

• Bumped the version requirement for mkdocs-ultralytics-plugin to >=0.0.38.
• Updated software version to 8.1.4.
• Added ASSETS to the __all__ export list in ultralytics/__init__.py.
• Changed manual download prompt behavior for the Argoverse dataset.
• Ensured class names are checked before use in ultralytics/engine/model.py.
• Fixed batch size derivation in ultralytics/nn/modules/head.py.
• Streamlined the build process for 2D sine-cosine position embedding in ultralytics/nn/modules/transformer.py.
• Improved Tensorboard callback functions with better exception handing and context-aware logging.
• Extended time limit for non_max_suppression operation in ultralytics/utils/ops.py.

🎯 Purpose & Impact

• Documentation plugin update ensures the latest features and fixes for the documentation generated by MkDocs.
• Version increment signifies a new iteration of the software with the incorporated improvements.
• ASSETS in __all__ export list allows users to access assets more conveniently from the Ultralytics package.
• Argoverse dataset download change addresses potential issues with automated downloads, guiding users towards a manual download.
• Class names check adds robustness to model loading, preventing potential crashes due to name errors.
• Batch size fix corrects a potential source of bugs in the head module, affecting how input is processed.
• 2D embedding calculation adjustments optimize performance and might resolve potential compatibility issues with different PyTorch versions.
• Tensorboard improvements enhance developer experience and provide better model graph visualization support, making debugging and analysis easier.
• Non_max_suppression time adjustment offers better performance by allotting more time for processing, especially useful for larger batch sizes.

[codecov]

Codecov Report

Attention: 6 lines in your changes are missing coverage. Please review.

Comparison is base (6535bcd) 77.16% compared to head (4976b50) 39.59%.

Files	Patch %	Lines
ultralytics/nn/modules/transformer.py	0.00%	3 Missing ⚠️
ultralytics/engine/model.py	0.00%	2 Missing ⚠️
ultralytics/nn/modules/head.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #7725       +/-   ##
===========================================
- Coverage   77.16%   39.59%   -37.57%     
===========================================
  Files         117      117               
  Lines       14336    14337        +1     
===========================================
- Hits        11062     5677     -5385     
- Misses       3274     8660     +5386     

Flag	Coverage Δ
Benchmarks	?
GPU	39.59% <40.00%> (+0.01%)	⬆️
Tests	?

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[glenn-jocher]