Use this project with recent GPU ?

[antoinersw]

Hello I have a 4080 (sm_89), according to the warning log CUDA is not compatible.

Is there a way to use a more recent version of CUDA/pytorch ?
I installed the project through the requirement package.

it results obviously by a :

Regards

[vladgohn]

same problem, no answers

[sebaturen]

In our case, we've opted for switching to a GPU compatible with CUDA11, specifically with version sm_75. While this approach is functional, it's not an optimal solution. Installing and configuring CUDA with a specific version can be challenging.

• we switch to GTX 20xx

[vladgohn]

I would generally like to run the project somehow, even on a CPU, just to make it work even slowly. do I have a chance to do this?

[sebaturen]

I don't think you can run it on a CPU; TensorFlow requires a GPU. For now, I managed to run the project by changing the GPU, but I'm having issues generating the landmarks.

[OchotonaPrinceps]

I managed to run it on an RTX4060.

It took me a couple hours to set up, but I am happy with the results.

[imxtx]

I managed to run it on an RTX4060.

It took me a couple hours to set up, but I am happy with the results.

Nice! Can you tell us how did you solve this problem?

[toseefkhan403]

I managed to run it on an RTX4060.

It took me a couple hours to set up, but I am happy with the results.

Hey! I'm stuck on the same problem, can you tell us how you got it working? Also, how long does the process usually take?

TIA 🐱

[urlan]

I managed to run it on an RTX4060.

It took me a couple hours to set up, but I am happy with the results.

Can you show us all modifications you have made? This project is old, so I'm quite sure it'll have a lot of problems to work properly.

[LauraCrx]

Can you show us all modifications you have made? This project is old, so I'm quite sure it'll have a lot of problems to work properly.

you have to make a conda env with a cuda-toolkit compatible with your card and then install a version of pytorch3d, torch and all other packages that are compatible with the cuda-toolkit your choice.

a exemples from another projetc:
pytorch3d-0.7.6 pytorch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 pytorch-cuda=12.1