Pangeo-binger using GPU

Education
1

Hello everyone !
I’m new here and I would like to have some information for a student project that I’m leading.

Me and my group have to create a webpage in which research teams could dynamically compute their codes, which includes deep learning techniques applied to oceanography. That’s why must give to the teams an access to a GPU. Without this constraint, PANGEO would have been our solution. Thus, we would like to know wether it would be possible to use the the Cloud computing service provided by Pangeo-binger using a GPU rather than a CPU.

I thank you for your attention, let me know if you need further information to answer my question. Sincerely,

Arnaud Remy

1 Like
2

Sounds exciting! It’s certainly possible to deploy GPUs on the Pangeo stack, though Pangeo binder doesn’t have it. I’d recommend reading this blog post on “Deep Learning with GPUs on Pangeo” at https://medium.com/pangeo/deep-learning-with-gpus-on-pangeo-9466e25bfd74 if you haven’t already :wink: Are you interested in setting up your own GPU Pangeo deployment, or using an off the shelf one?

3

@arnd_remy: thanks for this interesting question.

It is definitely possible to configure Pangeo with GPUs. This I believe this is set up now in our AWS deployment: http://pangeo.io/cloud.html#clusters

However, cloud GPUs cost significantly more money than CPUs, which is why they are not offered with binder. If you have a budget to pay for GPU computing for your team, there are lots of options.

  • You can run your own private Pangeo JupyterHub / Binder on your own cloud account
  • We could explore managed hosting options, possibly via 2i2c

If you don’t have a budget for computing and need a free GPU service, I think you only option is Google Collab.