I dont understand why my wandb_metadata.json file is showing this

[harrykwon97]

Hello, i am pretty new to this wandb function. I have been trying to run a program with it, but it shows an error of like this:
Traceback (most recent call last):
File “main.py”, line 105, in
main()
File “main.py”, line 99, in main
trainer.train(start_iteration=epoch)
File “/home/cs2212/Desktop/voxel2mesh-master/train.py”, line 58, in train
loss = self.training_step(data, start_iteration)
File “/home/cs2212/Desktop/voxel2mesh-master/train.py”, line 22, in training_step
loss, log = self.net.loss(data, epoch)
File “/home/cs2212/Desktop/voxel2mesh-master/model/voxel2mesh.py”, line 214, in loss
pred_points = sample_points_from_meshes(pred_mesh, 3000)
File “/home/cs2212/.local/lib/python3.8/site-packages/pytorch3d/ops/sample_points_from_meshes.py”, line 55, in sample_points_from_meshes
areas, _ = mesh_face_areas_normals(
File “/home/cs2212/.local/lib/python3.8/site-packages/pytorch3d/ops/mesh_face_areas_normals.py”, line 44, in forward
areas, normals = _C.face_areas_normals_forward(verts, faces)
RuntimeError: Not compiled with GPU support. (FaceAreasNormalsForward at /root/project/pytorch3d/csrc/face_areas_normals/face_areas_normals.h:51)

The preprocessing data job was done fine, but as I try to run the program with the preprocessed data, the upper error happens

I checked the metadata.json file and realized the cuda was set as null even though i checked the cuda was there with nvcc --version. I am guessing the wandb not realizing the cuda is there seems to be an issue. Are there any methods of how i could solve this? Any advice is appreciated Thank you

[ramit_goolry]

Hi @harrykwon97,

Looking at the traceback of this error, this issue does not seem to be originating from wandb. The wandb library does not communicate with CUDA directly, nor is CUDA nessecary for wandb to operate, so I would not expect it to break because of wandb.

It seems like the error actually originates from PyTorch3D, which would make sense since it would need to communicate to CUDA directly. I would suggest checking compatibility of your CUDA version with PyTorch 3D as a first step.

Thanks,
Ramit

[ramit_goolry]

​Hi @harrykwon97,

We wanted to follow up with you regarding your support request as we have not heard back from you. Please let us know if we can be of further assistance or if your issue has been resolved.

Best,

Weights & Biases

[ramit_goolry]

Hi @harrykwon97, since we have not heard back from you we are going to close this request. If you would like to re-open the conversation, please let us know!

​