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
I am wondering why the segmentation methods require spconv lib to have grad contiguous meanwhile some detection method based on spconv does not require this. Do you know what is causing problems?.
The text was updated successfully, but these errors were encountered:
Sry that I don't know the exact reason. If grad is not contiguous, the first stage converges well while the second stage fails. I guess the code written with CUDA (clustering and instance voxelization) makes the difference.
I am wondering why the segmentation methods require spconv lib to have grad contiguous meanwhile some detection method based on spconv does not require this. Do you know what is causing problems?.
The text was updated successfully, but these errors were encountered: