CPU version, Docker is running, but can't train/eval

[HamedMP]

The instruction for docker is for running the docker instance. But when you are inside the docker, you can't train/eval.

It's just like cloning the code, downloading all datasets, pre-trained models on your own computer.

OS: macOS

[brannondorsey]

I'm having trouble understanding your question. Are you getting specific errors when trying to run scripts from hanzhanggit/StackGAN w/ CPU only? If so, what errors are you getting specifically?

Best, Brannon

[HamedMP]

Thanks @brannondorsey for asking to clarification.

I will post stack trace later if needed, now a summary of what are the errors:

In either case of running code directly on the computer or using CPU-only docker container, the installation of torch libraries which depends on GPU fails. This results in not having full torch environment and then not being able to run get_embeddings.lua. The rest of the code will fail of course.

I wanted to transform the torch code to CPU only, but I don't have any practical experience in the torch. My guess is that the line requires GPU is:

txt = txt:cuda()

But I don't know what is the CPU counterpart :)

[brannondorsey]

Ugh oh, I'm afraid I might have similar skill limitations.

[brannondorsey]

If you post traces I can try and take a look.

[RJ722]

@HamedMP Were you able to run it successfully without GPU?

[apremkumar1989]

anyone successful without GPU?

I am getting the following error: module 'cunn' not found:No LuaRocks module found for cunn

[HamedMP]

Not at all, I found a GPU to use it.

Best Regards, Hamed MP http://bit.ly/athehamedmp* MSc student, Data Science | EIT Digital | +33 6 222 6 34 68

On Fri, Jun 16, 2017 at 6:32 PM, Rahul Jha [email protected] wrote:

@HamedMP https://github.com/hamedmp Were you able to run it successfully without GPU?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/brangerbriz/docker-StackGAN/issues/4#issuecomment-309072871, or mute the thread https://github.com/notifications/unsubscribe-auth/ADlMF_27XZomU0SIBZ6wZCcgBsizO-Utks5sEq4egaJpZM4M-bzx .