Can not load weights with Qt

I am trying to call pytorch code with C++, and the C++ code is called in Qt.

All the things goes right but when the code run:

weight = torch.load('./weights.pth')

Concretely, If i run the C++ and pytorch code without Qt, the program runs correctly. But if the code is implemented in Qt, the weights can not be loaded. By the way, random initialization is ok.

I am confused.
Could someone help me?

try

weight.load_state_dict(torch.load('./weights.pth'))

Thank for reply.

But it seems do not work. I think you mean that the code should be modified from

weight = torch.load('./weights')
model.load_state_dict(weight)

to

model.load_state_dict(torch.load('./weights'))

However, the core question is, PyTorch code is runable in pure Python environment, and then I write C++ interface to call the PyTorch codes. It is noraml and runable, there is no problem. However, in the next step, when I use Qt to call the C++ interface, the code (just ‘load’ command) does not work.

I don’t know is there some conflicts between Qt and PyTorch…