Maximum number of channels?

Hi all!

I trained two GAN models (the actual python scripts are copies of each other) - one using 7 and the other 9-channel images (stored as .tiff). When testing, the 7-channeled model works as expected, but the 9-channeled one returns all NaN values, even though during training the loss values were actual numbers. Is there a maximum number of channels PyTorch can handle?


No, PyTorch doesn’t set the max. number of channels to 9 and you can easily test it with a few unit tests.