Threaded TensorBoard graph logging

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Refined callback invocations and plot control in training routines.

📊 Key Changes

• Removed plots parameter from callback invocations in train.py.
• Added a self.plots instance variable in the Callbacks class to manage plotting.
• Simplified plotting-related code in Callbacks methods by using self.plots.
• Adjusted the thread decorator usage in the Logger's tensorboard graph logging function.

🎯 Purpose & Impact

• Enhances code maintainability by reducing the complexity of the callback functions' signatures.
• Provides a cleaner, more centralized way to control plotting, making it easier to enable or disable without affecting function parameters directly.
• Improvements in logging failures with more descriptive error messages in tensorboard graph logging offer users better debugging information.
• These changes aim to streamline training code in YOLOv5 and improve the user experience for those monitoring training progress. 👩‍💻📈