Notebook repo for learning Part 1 2019

fastuser · February 4, 2021, 7:31pm

Hi all,

I would like to learn and practice course videos found at https://course19.fast.ai/videos (i.e., Part 1 of the 2019 course).

Which Github repo do I find the relevant notebooks with runnable code in? As an example of how I tried, I found the below location:

github.com

fastai/fastai2/blob/master/dev_nbs/course/lesson6-rossmann.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai2.tabular.all import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Rossmann"
   ]
  },
  {
   "cell_type": "markdown",

This file has been truncated. show original

by looking at the address bar in the video http://course19.fast.ai/videos/?lesson=6

BUT looks like those notebooks are no longer available (“This repository has been archived by the owner. It is now read-only.” and getting a loading error from Github when trying to open that notebook even to just read).

FOUR questions (for any kind soul that knows about this):

Where can I access the notebooks for Part 1 2019?
Which version of the fastai library do those notebooks use?
Is that version stable without crashes/bugs etc?
If that version is deprecated, where can I find the equivalent but updated course material? (Part 1 of the 2020 course doesn’t cover Resnets, GANs, U-nets etc)

I tried reading through the forums and medium articles that (ex)students have created but the information is perplexing.

lukemshepherd · February 4, 2021, 10:03pm

Hey!

I’ve put the link to the 2019 course bellow but would strongly recommend doing the 2020 course (at least first).

The 2020 v4 course follows the chapters in fastbook which covers pretty much everything (including resents, GANs etc and more…)

But also, importantly, because the it uses the current version of fastai, if I remember correctly the 2019 used fastai v1. They look quite similar but the current version has some big improvements in how the dataloaders are constructed and far improved callbacks and you can use v1 code on the v2 library (and visa versa)

Hope that answers your question!

This is the 2019 repo, if the notebooks don’t load on github don’t worry sometimes they are a bit funny previewing! If you clone the repo and run them locally or on colab they will be fine.

fastuser · February 5, 2021, 10:33am

Hey Luke,

Thank you so much for your response. So it’s only a problem with the preview then. I will clone the repo like you suggested!!

I have completed the Part 1 of 2020 version but it doesn’t cover Resnets, Convolutions, GANs etc. So looking to learn those topics now.

Its true that fastbook covers pretty much everything. Would you know which 2020 course video lecture corresponds to this Convolutions chapter in Fastbook? Don’t trouble yourself too much if you don’t.

Thank you,
Sunny