Segmentation fault (core dumped)
Created by: Kracozebr
I've tried to run your code:
cd Feature_extraction
python main.py
But get segmentation fault.
Tried to debug and code crashes on this line:https://github.com/SMohammadi89/PointView-GCN/blob/200b7b2284951e2f2b1536ca673a72ba793bb474/Feature_extraction/pointnet_util.py#L37
I've checked size of tensors and tried to reproduce this error:
import torch
src = torch.randn(1, 512, 3)
dst = torch.randn(1, 1024, 3)
dst_permute = dst.permute(0, 2, 1)
# dist_permute = torch.randn(1, 3, 1024)
print("torch.matmul(src, dst_permute).size()")
print (torch.matmul(src, dst_permute).size())
dist = -2 * torch.matmul(src, dst.permute(0, 2, 1))
print("dist.size()")
print (dist.size())
src = torch.load("src.pt")
dst_permute = torch.load("dst_permute.pt")
dist = -2 * torch.matmul(src, dst_permute)
print("dist.size()")
print (dist.size())
The output:
torch.matmul(src, dst_permute).size()
torch.Size([1, 512, 1024])
dist.size()
torch.Size([1, 512, 1024])
dist.size()
torch.Size([1, 512, 1024])
But it works properly, I thought that it can be caused when matmul is processed with certain data, but I've stored the tensors and loaded them again to test it, but there is no error.
Can you help with this?