Open blackCmd opened 3 years ago
I solved it by running the below command when exporting to "torchscript": python export.py --weights yolov5l.pt --device 0 --inplace
But when I export with the above command, there is some performance depreciation. It means that the objects which are detected with pytorch detection function are not detected with C++ libtorch for the same image.
I don't know why.
I solved it by running the below command when exporting to "torchscript": python export.py --weights yolov5l.pt --device 0 --inplace
But when I export with the above command, there is some performance depreciation. It means that the objects which are detected with pytorch detection function are not detected with C++ libtorch for the same image.
I don't know why.
Wow Thank you. I wondering about the performance depreciation problem too. Does anyone know what is the cause?
Load and predict is OK. But when I try
.toTuple()
Error occured. How to solve it?