Help, "Segmentation fault" occur at "x = self.conv_input(input_sp_tensor)"

[drilistbox]

Thanks for your interesting work.

But "Segmentation fault" occur at "x = self.conv_input(input_sp_tensor)", i.e. "

MonoNeRD/mononerd/models/backbones_3d_lidar/spconv_backbone.py

Line 385 in 2d15e6f

x = self.conv_input(input_sp_tensor)

".

Waiting for your helps!!!

[cskkxjk]

I see similar problems in this issue.

Based on this spconv issue, the segfault could be attributed to the gcc version (traveller59/spconv#245 (comment)).

On Ubuntu 18.04, gcc-7.5.0 works for me. Please follow the instructions here carefully to install the spconv-1.2.1.

Besides, backbones_3d_lidar is only required to train the kitti models.
To test the kitti checkpoints or train waymo models, lidar models are not needed, so you can comment out LIDAR_MODEL in yaml configs. i.e.
"https://github.com/cskkxjk/MonoNeRD/blob/main/configs/mono/waymo_models/waymo_kitti_format.yaml"