Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Graphic card compatibility #147

Closed
aleksy opened this issue May 26, 2020 · 11 comments
Closed

Graphic card compatibility #147

aleksy opened this issue May 26, 2020 · 11 comments

Comments

@aleksy
Copy link

aleksy commented May 26, 2020

Hello there!

My question is if the program is suitable for PCs that only have Nvidia graphic cards, or do AMD support this program as well?
Yesterday I have encountered an error and it's about me not having Nvidia graphic card.
Neither native nor in docker I could make it run.

@AliaksandrSiarohin
Copy link
Owner

Yes, it need a cuda, which is nvidia library. You can use cpu or colab, if you don't have it.

@Zylvian
Copy link

Zylvian commented Jun 5, 2020

When using colab, where does the compiling actually take place?

@AliaksandrSiarohin
Copy link
Owner

Python doesn't need compilation.

@Zylvian
Copy link

Zylvian commented Jun 5, 2020

Sorry, I mean what resources does your Colab demo project use to make the video?

Also, here's something I made:
https://www.facebook.com/TheBestEmperor/posts/2827945487316041

@AliaksandrSiarohin
Copy link
Owner

This is an educational service from Google. So it uses Google gpus.

@Zylvian
Copy link

Zylvian commented Jun 5, 2020

Btw, is there a way to config the colab files to accept formats other than 256x256, or is this problematic?

And are there different checkpoints for different uses, or is it all the same which one you use?

@AliaksandrSiarohin
Copy link
Owner

AliaksandrSiarohin commented Jun 5, 2020

This is problematic and require retraining see related isssue #20.
Different checkpoints for different datasets (body, carton animals, etc). However for practical purposes only faces with vox-celeb checkpoint is useful currently.

P.S. Oblivion one of my favorite games.

@Zylvian
Copy link

Zylvian commented Jun 5, 2020

Haha, well I'm glad I could use your hard work for something entertaining then! :D

I'm working on getting a competent setup to test this with, but as I'm running on Windows, I'm running into a bunch of problems. Right now I'm trying to use anaconda to enable cuda support with pytorch, as running it with cpu was not scalable, and I keep getting "Torch not compiled with CUDA enabled".

Also, when it comes to faces, how much "room" do I have to test in? With using your Colab demo, the only real different stuff I can vary are the ``make_animation()` parameters, but from what I've gathered it's mostly either doing relative or absolute replacements.
If you have any tips, I'll gladly listen! :D

@AliaksandrSiarohin
Copy link
Owner

Sorry but I'm not familiar with windows instalations, so can't help you.
I only impliment the parameters in make_animation. But you can try to write a code for
keypoint manipulation yourself. For example fix position of some keypoint and move others, or some movements from one video and others from another video.

@Zylvian
Copy link

Zylvian commented Jun 5, 2020

For face transplants, i.e the Oblivion video, is it best to use taichi or taichi-adv?

@AliaksandrSiarohin
Copy link
Owner

AliaksandrSiarohin commented Jun 6, 2020

Taichi is not for faces. You should use vox or vox-adv. No idea which is better.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants