konioy
konioy
can it be understood as an augmentation of the GT box, up, or down, or right, or left to translate strid*0.5 pixels？
Is there a model for yolov5s-transform ？
Is there a pretrained model for yolov5s-transformer?
If you can't reproduce it, maybe it has something to do with the model. But the image size obtained by the above two methods is 736*551 and 736*552. The result...
You misunderstood what I was trying to say.
Is it a problem with the pytorch version? Is the 1.9 version of pytorch unreproducible?
I looked at (https://github.com/ultralytics/yolov5/pull/8213) and (https://pytorch.org/docs/stable/generated/torch.use_deterministic_algorithms.html?highlight=torch%20use_deterministic_algorithms#torch.use_deterministic_algorithms), but don't understand how torch.nn.Upsample is reproducible?
I am using pytorch1.9, which I know cannot be reproduced. It may be that the choice of convolution operator is different (because torch.cudnn, benchmark=true). But from the loss point of...
yes, i used --no-val. I think two repetitions of training, the effect on the validation set should be similar. But my experiments show that repeating the training twice, there is...
Thanks, I know. Do you have any suggestions for my situation?