Hi,
Genuine confusion here. I’m working through notebook #2 and learning the concepts, but I don’t feel like the course and/or book is geared towards gently bringing me into understanding the fast ai library. At this stage in the game, am I supposed to just sort of “click through” the code; or, am I supposed to be teaching myself the Fast AI library? Not sure what the intended workflow is here, could use some advice.
Thanks.
Hi delta_ark hope you are having a jolly day!
I have been using fastai since Feb 2019, and having used Tensorflow and Keras I believe this is as gentle as it gets. As the library is dealing with theories that have associated Phd papers, some of these papers took years to create.
I think the answer to your question above is yes to all of the question.
In the link below Jeremy describes how to learn fastai.
Learning fastai
The best way to get start with fastai (and deep learning) is to read the book, and complete the free course.
To see what’s possible with fastai, take a look at the Quick Start, which shows how to use around 5 lines of code to build an image classifier, an image segmentation model, a text sentiment model, a recommendation system, and a tabular model. For each of the applications, the code is much the same.
Read through the Tutorials to learn how to train your own models on your own datasets. Use the navigation sidebar to look through the fastai documentation. Every class, function, and method is documented here.
To learn about the design and motivation of the library, read the peer reviewed paper.
Having read 24K posts on this forum and created 455 I would add that everyone’s capabilities and learning skills are different.
My Personal workflow is:
I would say the most important concept for me is the dataloader concept. for me when I started to understand this I felt I was finally starting to get fastai.
There are many bright people on this site and at some time in their life, they have all been in your position.
One of these people is @muellerzr the person who wrote how he learns fastai. in this link. How do you... learn? (and study)
Having gone through the fast ai course and keeping up to date, I am confidently able to create certain models for business applications. (Still learning though).
Hope this helps.
mrfabulous1 
This was an incredible response. Thank you so much. OK, I just didn’t know what I was supposed to be attempting at this stage. Having an idea about the overall process people are using that seems to work is helpful. I wanted to dig into the library, but I wasn’t sure if this was the moment to do it; so, thank you. I actually also have a machine learning tutor, so we’ll probably go through these notebooks together.
You’ll be led deeper and deeper as you progress through the notebooks. It kinda starts out with “look at what you can do with ten lines of code” and moves on to “this is how it actually works” and then “this is how to tweak it and do different things with it”.
So how deep into the code do you recommend going at each notebook? e.g. at notebook 2 should I be going deep into the library, figuring out how a data loader and data block work, etc.
I’d say it depends who you are. If you have a solid programming background and feel like kicking it around every time they introduce a new bit, then by all means. Otherwise, I’d say the amount that you should be doing other than reading/running/understanding is probably the questions at the end of each chapter.