vp.decode ouputs NAN values... any help?

[IlkwonHong]

I am using cvpr19 branch, cuda10.0, pytorch 1.1.0, and windows11

I downloaded trained model and smplx model from smplify-x repo

For my case I cannot run any of given jupyter tutorial

from below code

#Sample a 32 dimentional vector from a Normal distribution
poZ_body_sample = torch.from_numpy(np.random.randn(1,32).astype(np.float32)).to('cuda')
pose_body = vp.decode(poZ_body_sample, output_type='aa').view(-1, 63)

print('poZ_body_sample.shape', poZ_body_sample.shape)
print('pose_body.shape', pose_body.shape)

images = render_smpl_params(bm, pose_body).reshape(1,1,1,400,400,3)
img = imagearray2file(images)
show_image(np.array(img)[0])

i got some nan values for pose_body as below

tensor([[    nan,     nan,     nan,  0.0290, -0.2194, -0.3330,  0.0783,  0.0997,
         -0.0150,     nan,     nan,     nan,     nan,     nan,     nan,     nan,
             nan,     nan, -0.0117,  0.1610, -0.0294,     nan,     nan,     nan,
             nan,     nan,     nan,     nan,     nan,     nan,     nan,     nan,
             nan,     nan,     nan,     nan,  0.0286,  0.0723, -0.4455, -0.0173,
          0.0512,  0.3359,     nan,     nan,     nan,  0.0081, -0.1747, -0.9129,
          0.0860,  0.3054,  0.8218, -0.0635, -0.8694,  0.3641,  0.0156,  1.1161,
         -0.2180,     nan,     nan,     nan,     nan,     nan,     nan]],
       device='cuda:0', grad_fn=<ViewBackward>)

what is wrong with this?

[SoftwareDevJake]

I'm having the same problem, any solution?

[IlkwonHong]

I updated the vposer with recent version at it worked

and you should change some code like

From

body_pose = vposer.decode(
pose_embedding, output_type='aa').view(1, -1) if use_vposer else None

To

body_pose = (vposer.decode(pose_embedding).get('pose_body')).reshape(1, -1) if use_vposer else None

[SoftwareDevJake]

What do you mean by undating to recent version? Do you mean I re-clone the git repository? and by the way, in the torchgeometry/core/conversions.py file, if i change the line 272, "mask_d2 = rmat_t[:, 2, 2] < eps" to "mask_d2 = rmat_t[:, 2, 2] > eps", it works for the vposer example, but when I try it in smplify-x, all the pictures get nan value from fitting stage 2. Have you tried it?

[IlkwonHong]

I'm not sure about what I've done to solve this problem.
What I remember is I got those errors when I cloned the cvpr19 branch.
I re-clone the repo with master branch, and downloaded ==VPoser v2.0== from https://smpl-x.is.tue.mpg.de/download.php
which will be used as --vposer_ckpt in smplify-x command args

I was so stressed solving this problem so I forgot most of the process

If this doesn't works for you I can send you my files

[SoftwareDevJake]

It'd be helpful if I check your code as well.
Can you send it to my email : [email protected]
And Thanks for you help.
Also, have you completed with smplify-x? Would you be able to help me since I'm stucked..?
If you're a Korean, we can make a phone call if you don't mind for easy comunication
Even if you are not a Korean, we can still communicate on the phone through snapchat if you don't mind

[IlkwonHong]

메일답변드렸습니다