Six chapters of the book now available for free

I’ve just made six of the most foundational chapters of the book available for free in a ready-to-read form. I hope you like them!


That’s great @jeremy. I can access the book using Colab but is there a way to access an interactive version via Jupyter? I followed the first couple of Live Coding videos which were a great help - managed to set up a LINUX environment on my Windows PC and have JupyterLab installed. Thank you

You can fork it via GitHub and run all of it locally if you have GPU. Colab is also interactive. Paperspace has free option too, you can run the book in the cloud GPU.


Hey, thanks @bahman_apl . I think I already forked it, but can’t seem to interact with it? Only read only?

In terms of GPU I have UHD Graphics 630

In order for things to run quickly you need an Nvidia GPU. Training on a CPU is extremely slow. Your best bet is to use Colab, Kaggle or Paperspace. Colab and Kaggle have a modified version of Jupyter and I believe you can use Jupyterlab on Paperspace. If you click the links in GitHub - fastai/fastbook: The fastai book, published as Jupyter Notebooks it will open in Colab which is interactive.

There is a getting started guide in Lesson 1 of the 2022 course here: Practical Deep Learning for Coders - 1: Getting started
You will also probably find these live coding sessions helpful: Live coding 1


Thank you Mat. I have been using Kaggle but it’s causing me major grief. I always end up with CPU usage of over 100%, sometimes as high as 225% (didn’t know that >100% was even possible), even when using their GPU processor. From review of the forums it seems like I’m not the only one having problems with Kaggle?

Maybe I’ll give Paperspace a go.

Yes, I have followed the first couple of Live Code videos which are awesome :slight_smile: I always thought LINUX was way over my head, but it’s really cool. I ran into one or two problems with installation but managed to fix. Will write a blog on it. Look like is the place. Fastpages seems to be out of favour?

The CPU % it’s reporting is an aggregate of each core which can be confusing. You definitely want to make sure that the runtime you have selected is GPU, not just CPU. I have not been running things in the cloud much recently so I’m probably not able to troubleshoot too much on that front. I have a dedicated workstation at home for AI work.

1 Like