search

vchoutas / smplx

SMPL-X
https://smpl-x.is.tue.mpg.de/
Other
1.78k stars 303 forks source link

`transfer_model` fails with CUDA error #111

Open gngdb opened 2 years ago

gngdb commented 2 years ago

It's not in the documentation but it seems like transfer_model.py requires a GPU? I get the following error trying to run it:

INFO - 2022-01-31 17:33:44,908 - acceleratesupport - No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate'
2022-01-31 17:33:45.081 | ERROR    | __main__:main:41 - CUDA is not available!

It doesn't print a stack trace so I don't know if there's an option to not use the GPU.

gngdb commented 2 years ago

Aha, it appears to be hardcoded in transfer_model/__main__.py. That's where the check happens and the error is generated: https://github.com/vchoutas/smplx/blob/master/transfer_model/__main__.py#L39-L42

gngdb commented 2 years ago

I tried disabled that hardcoded device setting and it works. It looks like it should take about 2.5 hours to process 601 .obj files.

gngdb commented 2 years ago

I set it to use CPU in my fork here but it's probably better to set this somewhere, for example in the config file.

Dipankar1997161 commented 10 months ago

It's not in the documentation but it seems like transfer_model.py requires a GPU? I get the following error trying to run it:

INFO - 2022-01-31 17:33:44,908 - acceleratesupport - No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate'
2022-01-31 17:33:45.081 | ERROR    | __main__:main:41 - CUDA is not available!

It doesn't print a stack trace so I don't know if there's an option to not use the GPU.

hey @gngdb,

I was also running the transfer model function,

I have seen the default folder for saving is at the root dir as output.. how can we pass it as a cmd argument??

I want to create a bash file for it since I have 100s of fitting to be done and I cannot put it on manually...

gngdb commented 10 months ago

Honestly, the easiest thing to do is going to be add a step in your bash script to move the output dir after each iteration so you don't have to edit the code. Good luck!