Immediate checkpoint serialization

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements in model checkpoint serialization for efficiency and speed 🚀.

📊 Key Changes

• Introduced io.BytesIO for in-memory checkpoint (ckpt) serialization before saving.
• Removed repeated metrics and results processing; now processed once and included in the serialization.
• Simplified checkpoint saving logic, enhancing code readability and maintainability.

🎯 Purpose & Impact

• Efficiency: By serializing the model checkpoint once and storing it in memory, the PR significantly reduces the time and disk I/O required during saving.
• Faster Training: This change minimizes disruptions during the training process, allowing for smoother and slightly faster training sessions.
• Cleaner Save Logic: The updated saving method simplifies future modifications and understanding of the code, beneficial for both maintainers and contributors.

Expect smoother training sessions with minimal save-related disruptions, a boon for users running large-scale model training. 🏋️‍♂️💾✨

[codecov]

Codecov Report

Attention: Patch coverage is 85.71429% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 76.94%. Comparing base (eed01a2) to head (c2a88b4).

Files	Patch %	Lines
ultralytics/engine/trainer.py	85.71%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #9437   +/-   ##
=======================================
  Coverage   76.94%   76.94%           
=======================================
  Files         117      117           
  Lines       14853    14854    +1     
=======================================
+ Hits        11429    11430    +1     
  Misses       3424     3424           

Flag	Coverage Δ
Benchmarks	36.73% <0.00%> (-0.01%)	⬇️
GPU	38.87% <85.71%> (+0.01%)	⬆️
Tests	72.07% <85.71%> (+<0.01%)	⬆️

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.