You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thank you for bringing this fantastic package to the community.
I inadvertently updated rmm and pylibraft-cu12 to version 24.10.0 and found that datasets with ~500k cells that I had previously had no problem running rsc.pp.neighbors on were consistently OOM'ng. When I downgraded those packages back to 24.2.0, I could run the datasets with no problem again. It seems that version 24.8.* also causes this OOM behavior. I am using rapids-single-cell version 0.10.10. Do you have any thoughts on this or suggestions as to which version is preferable in terms of accuracy of nearest neighbor graphs?
Thank you for your help and input!
The text was updated successfully, but these errors were encountered:
I haven't really looked at this and personally didn't encounter the issue yet. What GPU are you using? Can you reporduce the issue with the 500k testdataset?
Thank you for bringing this fantastic package to the community.
I inadvertently updated
rmm
andpylibraft-cu12
to version24.10.0
and found that datasets with ~500k cells that I had previously had no problem runningrsc.pp.neighbors
on were consistently OOM'ng. When I downgraded those packages back to24.2.0
, I could run the datasets with no problem again. It seems that version24.8.*
also causes this OOM behavior. I am using rapids-single-cell version0.10.10
. Do you have any thoughts on this or suggestions as to which version is preferable in terms of accuracy of nearest neighbor graphs?Thank you for your help and input!
The text was updated successfully, but these errors were encountered: