Kin-Yiu, Wong

Do your GPUs have tensor cores, if not, you have to set half to False to get correct results.

What is the AP you got of YOLOv7.

For AUC, you may need check if the precision and recall calculation are same or not. Usually, there are two kind of calculation, which are continuous and interpolation. Different calculation...

train_aux.py should use cfg/training/*.yaml.

Please use yolov7-d6_training.pt instead.

https://github.com/WongKinYiu/yolov7#transfer-learning

Could it run well on [u5 branch](https://github.com/WongKinYiu/yolov7/tree/u5)?

It mainly because we use [iou as target of objectness](https://github.com/WongKinYiu/yolov7/blob/main/utils/loss.py#L610), you could reduce self.gr to avoid this issue. By the way, my testing result of yolov7 is as below. 

Yes, to handle this case, the model need to be retrain by above mentioned suggestion.

No, self.gr is set in [train.py](https://github.com/WongKinYiu/yolov7/blob/main/train.py#L288).