Expected object of type torch.FloatTensor but found type torch.cuda.FloatTensor about super-slomo HOT 6 CLOSED

@Shivanshmundra This error normally occurs when the device here and here is "cpu" while at other places it is "cuda:0" / gpu. Basically the flow back warper is on cpu while the tensor is on gpu. How are you changing the code to run on CPU/GPU. Did you make any modification to the code?

from super-slomo.

@avinashpaliwal I just changed device = "cpu" so that PyTorch uses cpu forcefully in code here.
I tried exporting model and data to cuda device after they were loaded but still, same error occurred.

from super-slomo.

When you are manually exporting the model try to set this and the device on the next line to cuda as well.

from super-slomo.

I used the pre-trained model given by you and changed the variable device value to cpu/cuda:0. So Every line now should have used the value of device variable and hence should not be a problem due to that.
If you can once try to convert any video to slowmo by pretrained model from latest commit in this branch, it would imply that I am making mistake somewhere.

from super-slomo.

I have tested it and it's working fine. My configuration is: PyTorch 0.4.1, python 3.6.7, Cuda 9.2, Windows. I have not tested it with other configurations and therefore can't say for sure about the issues that will crop up.

• Was the script working for you at some commit version?
• Did you also try running from a fresh clone/fork?
• If some version of the code was working, then take a diff between the two versions of scripts (maybe also other dependent files) and make incremental changes to the working code to find out where the problem is.

from super-slomo.

I am assuming that your issue was addressed and closing this issue.

from super-slomo.