compute_on_cpu=True not respected; is it anomalib's fault?
#697
Unanswered
jpcbertoldo
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
So I had this issue Lightning-AI/metrics/issues/1323, and I don't know if it's rather on
torchmetrics's side oranomalib's.TLDR: passing
compute_on_cpu=Truetoanomalib.utils.metrics.auroc.AUROCcauses the tensors to be kept in the CPU during the updates but they are somehow transferred to the GPU atcompute().I supposed it is in
torchmetricsso I created the issue there.Anyone one has a better guess?
Beta Was this translation helpful? Give feedback.
All reactions