LoGG3D_net Checkpoint mismatch
uljad opened this issue · 3 comments
uljad commented
Thank you for the great dataset and the documentation. I am trying to replicate the LoGG3D evaluation but when I load the checkpoint there appear to be size mismathces that cause the following RunTime Errors:
RuntimeError: Error(s) in loading state_dict for LOGG3D:
size mismatch for spvcnn.classifier.0.weight: copying a param with shape torch.Size([32, 61]) from checkpoint, the shape in current model is torch.Size([16, 61]).
size mismatch for spvcnn.classifier.0.bias: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([16]).
size mismatch for sop.fc1.weight: copying a param with shape torch.Size([512, 1024]) from checkpoint, the shape in current model is torch.Size([128, 256]).
size mismatch for sop.fc1.bias: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([128]).
size mismatch for sop.fc2.weight: copying a param with shape torch.Size([256, 512]) from checkpoint, the shape in current model is torch.Size([64, 128]).
size mismatch for sop.fc2.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([64]).
peymmo commented
@kavisha725 could you please address this comment.
kavisha725 commented
Thanks for your interest.
Please check the latest commit in the LoGG3D repository.
To load the wild-places checkpoint, change the feature_dim
parameter here to 32 when you initialize the LoGG3D model.
uljadberdica1000 commented
Yes I did that and it worked. Thank you!