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 trying to implement the FlowNet3D used for kitti. As in here: FlowNet3D
I am running into an error I can't get around:
File "/home/zuanazzi/PointCloudFlow/flownet3d_pytorch/util.py", line 396, in forward
dists[dists < 1e-10] = 1e-10
RuntimeError: copy_if failed to synchronize: cudaErrorIllegalAddress: an illegal memory access was encountered.
from flownet3d_pytorch.util import PointNetFeaturePropogation
pc1 = torch.zeros(2, 3, 8192)
l1_pc1 = torch.zeros(2, 3, 8192)
feature1 = None
l1_fnew1 = torch.zeros(2, 256, 8192)
layer = PointNetFeaturePropogation(in_channel=256, mlp=[256, 256])
out = layer(pc1, l1_pc1, feature1, l1_fnew1)
print(out.shape)```
I am using a v100-32 gpu. The model implemented in this repo runs fine, but it fails when I change the parameters of the layers.
Any ideas on how to fix it?
The text was updated successfully, but these errors were encountered:
Hello,
I am trying to implement the FlowNet3D used for kitti. As in here: FlowNet3D
I am running into an error I can't get around:
The text was updated successfully, but these errors were encountered: