Fastai book Deep Learning for Coders with fastai and Pytorch

description of my learning environment, following fastai reco
fastai
jupyter
fastbook
Published

September 24, 2020

Paper version of fastai book

Of course the 1st tep is to purchase this great book:

alt text

I have liked what Jeremy Howard said about why this is important to purchase it (in video 1). Fastai is offering full access to the book as notebooks. So that we can run all codes from them.

Get notebook version of fastai book

~/git/guillaume$ git clone https://github.com/fastai/fastbook.git

alt text

I have now a perfect combo between paper book and notebooks.

Video courses based on fastai book

Rachel Thomas and Jeremy Howard have done some great videos about learning fastai (and pytorch).

They tend to do it every year. But this year is quite special due to fastai book.

Here are all the 1st 7 videos: https://course.fast.ai/videos/?lesson=1

Fastai forums

This is the natural source of information and interactions with other students.

There is a category Part 1 (2020) which seems perfect: https://forums.fast.ai/c/part1-v4/46

Personal git organization

~/git/guillaume$ ll
fastai/                 <-- from https://github.com/fastai/fastai 
fastai_experiments/     <-- I will keep all experiments I will do here
fastbook/               <-- from https://github.com/fastai/fastbook.git 
guillaume_blog/         <-- from git@github.com:castorfou/guillaume_blog.git

fastai_experiments likely to have 1 notebook per chapter or video.

update jupyter to include extensions (toc, …)

conda install -c conda-forge jupyter_contrib_nbextensions

alt text I like table of content, others are quite usefull as well (scratchpad, ExecuteTime…).

Install some libraries to run book examples

conda install -c fastai fastbook

It will install graphviz, nbdev, and other libraries.

And most of them can be loaded by calling from utils import *

Launch jupyter notebook and start expermenting

cd ~/git/guillaume
conda activate fastai
jupyter notebook

And launch several tabs at: blog entries, fastai experiments, fastai courses and fastai videos.

And I keep track of progress with git.