crtie/Leveraging-SE-3-Equivariance-for-Learning-3D-Geometric-Shape-Assembly

BUG: File "/BreakingBad/multi_part_assembly/models/vnn/network.py", line 209, in forward global_inv_feats.reshape( RuntimeError: shape '[-1, 1024, 1024]' is invalid for input of size 276480

Closed this issue · 3 comments

wzm2256 commented

Hi, thanks for sharing your code.

I'm running the script on breaking bad dataset and getting this error. Any idea how to resolve this?

By the way, the set_trace in line 56 in the same file should be comment out, otherwise the code will break in that line. Is this the final version of your code?

wzm2256 commented

The error is caused by reshaping a tensor of shape (batch_size, num_parts, 2*feat_dim, 3) to (batch_size, 2*feat_dim, 2*feat_dim).

warshallrho commented

Hi, thanks for sharing your code.

I'm running the script on breaking bad dataset and getting this error. Any idea how to resolve this?

By the way, the set_trace in line 56 in the same file should be comment out, otherwise the code will break in that line. Is this the final version of your code?

For the set_trace issue, the released code will not run into those lines.

crtie commented

This was an oversight when we released the code, which has been corrected.