How do you get a GPU machine type in Codespaces?

[WellRestedSloth]

Select Topic Area

Question

Body

How does one get a GPU machine type in Codespaces?

I filled out the "SurveyMonkey" survey a couple of weeks ago, asking and requesting a GPU type.

I am on the free account type. I cannot find the documentation that says you need another account type for GPU. Am I correct?

I would like this (GPU type) so I can learn CUDA C++ programming. Thank you.

[jge162]

Hello @WellRestedSloth Codespaces, which is a GitHub feature, currently does not support GPU machine types on the free accounts.

[jge162]

there are many options out there, Google Cloud Platform (GCP) or Amazon Web Services (AWS) that allows you to rent GPU-enabled virtual machines. hope that answers your question. bests to u

[WellRestedSloth]

Thank you for your reply.
Can you point me to the documentation where it says that free account does not support GPU in codespaces? Or where GPU would be supported with a non-free account?

I see you have a Pro account. Can you confirm that the GPU type is available to you? Thanks.

[WellRestedSloth]

I did what you said, except the green button "New Codespace" page, there is no "custom" option or "Environment" section. Next to "Repository", I searched for "CUDA-Toolkit" and found "NVIDIA/cuda-samples". The Machine-type would still not show GPU. Even searching for "deep learning", no matter what repository I selected, I could not get GPU type.

Can you confirm that you are able to get a GPU-type? It appears you have Github free account like myself. I don't know if there's some limitation because of that.

[fretchen]

same here. Did someone find a way to have access to GPU machines ?

[WellRestedSloth]

Nope. I still don't and I haven't heard anything from Github.
In the meantime I have found "Saturn Cloud" dot io. They have GPU instances for free, 30 hours a month. From my testing, once you stop the instance your files are still there but any packages need to be reinstlled (i.e. gdb, build-essential, nvidia-cuda-toolkit, nvidia-cuda-toolkit-gcc). I've installed the CUDA extensions into VSCode and can step and debug CUDA apps. Not sure if that is what Saturn Cloud is intended for (I think it's more for machine learning, Pytorch, Jupyter, etc.) but so far works for what I want.

[naga-k]

I think this might be a AI generated answer.

[aosama]

hello Github team,

first my compliments on such a very well put together services.

I have a PRO account, is it possible to launch a GPU machine within Codespaces?

I do some ML related experiments inside vscode in Jupyter notebook and would be great if I can launch a GPU instance in Codespaces and run an experiment.

thanks
Ahmed

[man-chi]

@github team,
same here. how can I apply for a GPU codespace instance? what kind of application and account is required?

[github-actions[bot]]

🕒 Discussion Activity Reminder 🕒

This Discussion has been labeled as dormant by an automated system for having no activity in the last 60 days. Please consider one the following actions:

1️⃣ Close as Out of Date: If the topic is no longer relevant, close the Discussion as out of date at the bottom of the page.

2️⃣ Provide More Information: Share additional details or context — or let the community know if you've found a solution on your own.

3️⃣ Mark a Reply as Answer: If your question has been answered by a reply, mark the most helpful reply as the solution.

Note: This dormant notification will only apply to Discussions with the Question label. To learn more, see our recent announcement.

Thank you for helping bring this Discussion to a resolution! 💬

[sonam-shrish]

We need GPU for codespace please!