Implement @torch.no_grad() decorator

[glenn-jocher]

Replace with torch.no_grad(): statement with @torch.no_grad() decorator.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancement of the inference and testing code in the YOLOv5 project for better readability and performance.

📊 Key Changes

• Applied the @torch.no_grad() decorator to the detect() and test() functions in detect.py and test.py.
• Removed nested with torch.no_grad(): blocks inside these functions.
• Simplified conditional logic in detect.py for updating models.

🎯 Purpose & Impact

• 🚀 Increases efficiency: Disabling gradient computations globally with the @torch.no_grad() decorator to avoid redundancy.
• 📖 Improves code readability: By removing nested with torch.no_grad():, the code structure is simplified making it easier to read and understand.
• 🛠️ Enhances maintainability: Simplified conditionals reduce complexity, making it easier to maintain and update the code.
• 🌍 User experience: These changes aim to streamline the inference and test processes, potentially improving execution time and lowering resource consumption, which benefits all users.