RuntimeError when converting image to tensor about ssd.pytorch

[ariangilesgarcia]

Ok sorry for leaving an issue before trying everything I could to fix it.
I don't use torch so I had no idea what to do, but here's how I fixed it just in case.

On data_augment.py I had to chage return torch.Tensor(img) to return torch.FloatTensor(img).

Then an error appear saying it was expecting a CPU tensor and got a CUDA tensor instead, so I had to add .cuda() to the input variables (I don't know if there is another way to do it, I don't really know how to use pytorch yet). For example, changing x = Variable(transform(img).unsqueeze(0)) to x = Variable(transform(img).unsqueeze(0).cuda())

Again, sorry for the issue!

[amdegroot]

Ah, thanks for pointing that out, it must only be the case on GPU. For some reason I was using torch.Tensor() instead of torch.from_numpy() in data_augment.py. It should all be fixed now. Feel free to send a Pull Request if you find any other errors!