@AlexeyAB @WongKinYiu I would like to know which cuda version and pytorch version are used to produce output for YOLOv7-E6E model as reported in the paper? And specifically with --batch-size 128 using the following command:

train p6 models

python -m torch.distributed.launch --nproc_per_node 8 --master_port 9527 train_aux.py --workers 8 --device 0,1,2,3,4,5,6,7 --sync-bn --batch-size 128 --data data/coco.yaml --img 1280 1280 --cfg cfg/training/yolov7-e6e.yaml--weights '' --name yolov7-e6e --hyp data/hyp.scratch.p6.yaml

I am trying to reproduce the results for YOLOv7-E6E model reported in the paper.