Yes, deep learning is to use GPU and large dataset to train large models, therefore it is reasonable for all the notebooks to assume GPU availability and huge dataset storage space.
However, there are still billions of people having no free access to GPU and large storage for big dataset, nor they can afford physical GPU and large storage.
Fast.ai is very popular and famous, people around the world want to make the most out of it by learning all the coding skills, practical knowledge and theory of deep learning through experimenting the notebooks. For people who are denied with free access to GPU and large dataset storage, they may still leverage the most out of fast.ai notebooks if a very tiny version of the datasets are provided.
A tiny version of datasets provided on github has the following benefits
easy to download
enough space on local computer to handle
notebooks can run perfectly neither original large dataset nor GPU
models and source code can be experimented and learned thoroughly with only CPU
In this way the less privileged can still make the most out of fast.ai.
Hi @Daniel
Did you get a chance to checkout the Colab section on the course website?
They offer free GPU instances and there is a walkthrough of how-to set everything up.
Please check the forums for a discussion thread on it if you have any issues setting it up.
Hi @Payback, thanks for being curious, but I don’t want to name it, and you know there is only one country on this planet big enough to have billions of people denied access to google. I guess India is bigger but has open access to google.
Hi @Daniel
I understand what lengths you’re going to in order to learn
You can also try services which provide credits based on referrals (like Paperspace) or sign-up for services which provide abundant credits to students (AWS Educate, Azure Students, etc.)
Also, to get the ball rolling you can start with a referral chain with all your fellow students. On Paperspace you’ll get $10 with this link.