Closed lguerard closed 1 year ago
Great point. I think I can do a version check of pytorch before calling that function to avoid that issue.
Ok, latest version (including pip versions of omnipose
and cellpose-omni
) have fixed this issue. I just use a try/except branch to catch it for any version lacking backends
.
Hi,
I have been having issue with omnipose in the last couple of days and found out that the
mps
support in your latest version of Cellpose is giving issues with old version ofpytorch
(1.8.2) as it doesn't havebackends
. Would it be possible to bypass this for olderpytorch
version ?Thanks !