CGAN with images as labels?


I’m following a CGAN tutorail by Aladdin Persson. However, since I’m working with satellite data, I have no labels per se. In lieu of labels, I have “constrained images”: they corresponding images (dupliactes) with some of the pixels removed (let’s say 5% of pixels removed, so that the CGAN doesn’t alter the 95% of the remaining pixels, just the missing ones). All images are 2 channel (grayscale and alpha).

I’ve managed to edit the dataset to accept images as labels. However, upon executing the code, it returns

RuntimeError: Expected tensor for argument #1 ‘indices’ to have one of the following scalar types: Long, Int; but got torch.FloatTensor instead (while checking arguments for embedding)

Is there a remedy for this? Could you recommend me any other solution or perhaps a book/tutorial that expands on the embedding part of the code?