Fast.ai Setup for Google Colaboratory

I dropped into a study group today that was trying to work through the first Fast.ai course. It was the first day of the group, so most of the day was just spent getting everyone to the point where they could start working on the notebooks. I have my own setup available for Keras and Tensorflow on my home machine, but the fast.ai project wants Torch, and we needed something that everyone could get up and running quickly.

I spent most of my time getting the environment set up on Google's Colaboratory so that we could get people started without needing any local setup (just their Google account). This worked really well once I realized the data-set is not being downloaded from Kaggle, so you don't need to muck about with getting a Kaggle API key to get started. I've shared the initial-setup Colaboratory Notebook on Google Docs. Just click the link, and when asked, open the file with the Colaboratory drive app. Another attendee also created a docker instance that allows for CPU-only operation, but includes the dogs-and-cats dataset already.

I doubt I'll continue on with this study group, I was thinking it was a "study and implement the latest/most interesting papers in machine learning" type study group, where it's more of a "get an introduction to machine learning" group.

Comments

No comments yet.

Pingbacks

Pingbacks are closed.

Post your comment