GFLOPS Update

[glenn-jocher]

GFLOPS update to resolve issue #1751. GFLOPS now displayed at 640x640 inference size (TTA at 832) for fused models.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Updated performance metrics from FLOPS to GFLOPS in YOLOv5 README for clarity.

📊 Key Changes

• FLOPS in the pretrained checkpoints table have been changed to GFLOPS.
• Numerical values for FLOPS have been updated to reflect the new GFLOPS units.

🎯 Purpose & Impact

• Enhanced Readability: Converting FLOPS to GFLOPS simplifies the comprehension of computational performance for users.
• Improved Understanding: The use of GFLOPS, a more commonly used unit, helps users better understand the model's efficiency and compare it with other models.
• Potential to Inform Decisions: Users can make more informed decisions about which model to use based on their hardware capabilities and needs for speed and accuracy.