Where is URLs.PETS defined?

ml22 · August 31, 2020, 5:44pm

Hi all,

I’m looking at Course v3 - 2019, Lesson 1.

The Jupyter notebook is here:

fastai/course-v3/blob/master/nbs/dl1/00_notebook_tutorial.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Important note:** You should always work on a duplicate of the course notebook. On the page you used to open this, tick the box next to the name of the notebook and click duplicate to easily create a new version of this notebook.\n",
    "\n",
    "You will get errors each time you try to update your course repository if you don't do this, and your changes will end up being erased by the original course version."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Welcome to Jupyter Notebooks!"
   ]
  },
  {
   "cell_type": "markdown",

This file has been truncated. show original

In the notebook, URLs.PETS is defined from the start.

Before doing anything, if you run URLs.PETS
you get output.

URLs.PETS
‘https://s3.amazonaws.com/fast-ai-imageclas/oxford-iiit-pet’

I’m used to seeing variables defined before they are used.

Where is URLs.PETS defined?
Are there other variables?

Thanks

ml22 · August 31, 2020, 5:45pm

Image here:

URL_PETS

ml22 · August 31, 2020, 5:59pm

Ok, it seems to be part of the module:
fastai.datasets

PETS: A 37 category pet dataset with roughly 200 images for each class.

Am I correct on this?

JonathanSum · August 31, 2020, 6:12pm

It is part of the framework.
You can use it to download the data.

path = untar_data(URLs.PETS);
After that, you will download it into the path location.

ml22 · August 31, 2020, 8:25pm

Thanks

I’ve always seen:

Data, and
Code

In addition to the rest of the tech stack

Here, you don’t even need to
download the dataset. The code, or
objects do it for you.

This is a new paradigm.