Since torchvision is not compatible with Pytorch version 0.1.12, I am running this project with torch==1.1.0. And the error occur in line 119. Adding keepdim=True to torch.normcould prevent from that.
RuntimeError: The expanded size of the tensor (64) must match the existing size (66) at non-singleton dimension 2. Target sizes: [256, 66, 64]. Tensor sizes: [256, 66]
Since torchvision is not compatible with Pytorch version 0.1.12, I am running this project with torch==1.1.0. And the error occur in line 119. Adding
keepdim=True
totorch.norm
could prevent from that.