Update models/hub/*.yaml files for v6.0n release

[glenn-jocher]

Updates all extra YOLOv5 models in the models/hub directory to v6.0 release architecture.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Upgraded YOLOv5 configurations to v6.0, introducing model structure enhancements and feature integrations.

📊 Key Changes

• Updated backbone and head definitions across various YOLOv5 model configurations (BiFPN, FPN, P2, P6, P7, PANet, Ghost, and Transformer versions).
• Replaced Focus layers with Conv layers with updated arguments.
• Adjusted the number of C3 layers in backbone configurations.
• Swapped SPP (Spatial Pyramid Pooling) layers with the more efficient SPPF layers.
• Marked concatenations specifically within the BiFPN configuration.
• Updated the Transformer (C3TR) layer configuration.
• Added functionality to test all YOLOv5 models via a command-line argument.

🎯 Purpose & Impact

• The transition from Focus to Conv layers is expected to standardize the initial convolution process across various models.
• Reducing the C3 layers might optimize computational efficiency without significantly affecting model accuracy.
• The switch to SPPF should streamline spatial pyramid pooling while preserving feature diversity with reduced computational overhead.
• Explicitly marking concatenation steps enhances code readability and could help in debugging the BiFPN model.
• Refining the Transformer model aligns with updates in transformer technology and potentially boosts performance.
• The added test argument (--test) simplifies the process of verification for all model configurations, ensuring robustness before deployment.
• These changes aim to improve the models' performance, resource utilization, and maintainability, which benefits users with more efficient and effective object detection tools. 🚀🧠

🛠️ Overall, the PR represents a stride towards evolving and optimizing the YOLOv5 architecture. Users might benefit from faster, sleeker, and possibly more accurate object detection in applications.