Enhance argument loading when resuming. Fixes #6553

[R-N]

Issue #6553, resume=True doesn't actually enforce caching. It was caused by resuming a model that was trained with caching. This PR allows overriding cache and epochs. It also removes and renames loaded arguments that are now deprecated.

🤖[deprecated] Generated by Copilot at dd72f57

Summary

🔄🗑️🐛

Fix check_resume function in trainer.py to handle outdated arguments in checkpoint files. This improves compatibility and error handling when resuming training.

check_resume adapts to old or renamed args fall leaves change their hues

Walkthrough

• Update check_resume function to handle obsolete, renamed, or negated arguments in checkpoint file (link)

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

📊 Key Changes

• Added an exclude_class option in plot to ignore specific classes when plotting results.
• Updated argument handling in check_resume to clean up obsolete arguments and accommodate renamed or negated arguments.

🎯 Purpose & Impact

• The exclude_class option helps users to focus on relevant detection classes by excluding unwanted ones, enhancing the clarity of visual results. 🎨
• Improved argument management when resuming training ensures compatibility and reduces memory usage to prevent CUDA Out Of Memory (OOM) crashes. This increases the robustness of model training. 🛠️

🌟 Summary

Now you can exclude certain classes from your detection plots, and resuming model training is smoother and less prone to memory issues! 🚀

[github-actions[bot]]

CLA Assistant Lite bot:
Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request}

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 57.14286% with 6 lines in your changes missing coverage. Please review.

Project coverage is 79.90%. Comparing base (bc5b528) to head (90b88d1). Report is 856 commits behind head on main.

Files with missing lines	Patch %	Lines
ultralytics/engine/trainer.py	58.33%	5 Missing :warning:
ultralytics/engine/results.py	50.00%	1 Missing :warning:

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6854      +/-   ##
==========================================
- Coverage   82.24%   79.90%   -2.34%     
==========================================
  Files         106      106              
  Lines       13023    13036      +13     
==========================================
- Hits        10711    10417     -294     
- Misses       2312     2619     +307     

Flag	Coverage Δ
Benchmarks	37.75% <0.00%> (-0.04%)	:arrow_down:
GPU	41.46% <0.00%> (-0.05%)	:arrow_down:
Tests	74.62% <57.14%> (-2.94%)	:arrow_down:

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

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[Kayzwer]

@R-N you have to comment this "I have read the CLA Document and I sign the CLA"