Issue running demo script

[nshreyasvi]

Hello,
Thank you for your great work. I was trying to run the demo script but was getting the following error:
demo/input/000000061418.png libcudart.so.10.1: cannot open shared object file: No such file or directory
I have installed cuda toolkit as well as pytorch with cuda. When I used nvcc --version it shows the 11.8 and also when I do torch.version.cuda and torch.cuda.is_available() which returns True.
Could you let me know how to resolve this issue?

[karShetty]

This means your torch is running on Cuda 10.1 but you have CUDA 11.8 installed. You could try reinstalling torch in a new venv for any specific torch version / cuda toolkit

[nshreyasvi]

Hello, Thank you so much, I managed to solve the issue but when I run the demo script I get the following error

Expected size for first two dimensions of batch2 tensor to be: [1, 40] but got: [30, 40].

I use the command:

python demo.py --img_dir demo/input/000000296657.png --out_dir demo/output/ --checkpoint checkpoint_sparse.pt --model MeshRegSparse

Could you let me know what might be the issue here?

[nshreyasvi]

Hello, sorry for the delays. I was able to resolve the issues and render the output meshes. I changed line 130 in pliks/utils.py to the following:

def vtk_write(path, polydata, type='obj'):

and uncommented line 157 and 158 changing them as

vtk_write(os.path.join(self.options.out_dir, out_name + '_res_%02d.obj') % (_i,),
          create_polydata(pred_vK[0].cpu().data.numpy(), self.smpl.faces.cpu().data.numpy()))

so that I can save the output .obj files.

but would like to know how can I save the output in .pkl format with global_pose and other keys?