support specfiy multiple frozen layers

[youyuxiansen]

related to 6001

@glenn-jocher HI, I'm here. Now is

python train.py --freeze 10  # freeze up to 10
python train.py --freeze 5 6 7 8 9 10  # freeze layers 5-10
python train.py --freeze 5 7 8 10  # freeze layers 5,7,8,10

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced layer freezing flexibility during model training in YOLOv5.

📊 Key Changes

• Modified the freeze behavior in train.py to support freezing specific layers.
• Updated argument parsing to accept multiple values for the --freeze flag.

🎯 Purpose & Impact

• 🎯 Purpose: Allows more granular control over which layers of the neural network are frozen during training, enabling users to fine-tune their models more precisely.
• 💥 Impact: Users can now choose specific layers to freeze instead of only being able to freeze up to a certain layer. This could improve model performance and training efficiency for various use cases, particularly when fine-tuning pre-trained models.