On Monday, May 28, 2018, kirby urner <kirby.urner@gmail.com> wrote:

Wes or someone may have linked to this already. Just tuned it in myself:


CoLab is pretty cool. They have GPU instances.
Kaggle also has GPU instances now with Kernels and Learn.
https://www.kaggle.com/kernels
https://www.kaggle.com/learn/overview

We worked on a Kaggle data science competition as a team of individuals teaching each other through our local Python Users Group. It was a house prices prediction competition; similar to the well-known Boston house prices dataset included with scikit-learn.

There are a bunch of hosted Jupyter Notebook services now:
https://github.com/markusschanta/awesome-jupyter/blob/master/README.md#hosted-notebook-solutions

To host local instances of Jupyter for a group of size n, there are JupyterHub 'spawners' and 'authenticators'. 

https://zero-to-jupyterhub.readthedocs.io/en/latest/

https://github.com/jupyterhub/jupyterhub/wiki/Spawners

https://github.com/jupyterhub/jupyterhub/wiki/Authenticators


Gvisor is strongly recommended for sandboxing hosted containers with Docker:
https://github.com/google/gvisor


Binder builds upon JupyterHub but doesn't have auth yet:
https://github.com/jupyterhub/binderhub/issues/323
 

That's Google's way of letting us use Jupyter Notebooks in the cloud and to share them on Google Drive.

Storage:
https://github.com/jupyter/jupyter-drive

Storage + Real-time collaboration:
https://github.com/jupyterlab/jupyterlab-google-drive
 

I see where students would benefit, not that this is the first or only cloud-based environment.  Another great tool.

Kirby