Open adler-j opened 5 years ago
It should be noted that there are other options, e.g. azure notebooks but these don't provide a GPU.
So what's the TODO here for us? To make an example and link to it from the README? As some kind of inspiration? Or do you envision a more in-depth mention in the docs? As an alternative to installation, for instance.
I think we could add this to the installation doc, as an alternative to conda. Perhaps we could somehow also have a ready made notebook so it's basically click and run for our users.
There is now a working version here:
https://colab.research.google.com/drive/1wsHNyuK0GAM8sh9kBTzwh8t6Kuivj1KW
It's quite easy, so we should perhaps add it to the README (e.g. "click here to try".)
Google colab is an online jupyter notebook-like service that runs in the cloud and allows users free access to a GPU. For obvious purposes this is quite useful as a learning tool and to let users access code that needs a GPU even when they don't personally own one.
It seems that pip installable libraries can be trivially installed, see e.g.
Conda is harder https://github.com/astra-toolbox/astra-toolbox/issues/194 but could probably be fixed.
If we could make a simple example of this it would make introducing ODL in e.g. courses far easier.