So, I have an issue with some code where GPU:0 always seems to have abnormally high usage (and it gets worse with bigger inputs). Is it possible to see what tensors and models exist on a particular GPU? And if so, how would I do that? The hope is that I could use this information to help determine the cause of the issue.
Note that if your
tensor are on GPU, you will get a output, otherwise it will through and error. They will return the
id of the currently being used GPUs.