Thread by @TheZachMueller

Why does #nbdev do such weird namings for your notebook, such as "00_core.ipynb?"

There's actually a few reasons. Let's talk about that 🧵

First, it helps keep things organized module wise. Having everything numerical let's you section off by groups how certain segments of code are laid out.

An example of this is in @fastdotai, where notebooks starting with 20 are generally vision tutorials

But there's ~actually~ a second reason why this can be super cool!

In GitHub, currently when we run the tests for our notebooks, we run them all at once through calling `nbdev_test_nbs`. But we can actually speed this up by calling ~groups~ of notebooks! How does this work?

GitHub actions has a concept called "runners." Where essentially I can group together certain parts to run in different environments and customizations on their own, to speed things up. And these are controlled through a "matrix". An example can be seen here:

So what winds up happening is we split up one job into roughly 8 to 12 jobs for just the notebooks, as only small sections of notebooks get tested at a time.

This both allows us to isolate problematic areas in the library, ~and~ drastically speed up the time it takes!

To use this, we just modify the "testing" section of our GitHub action to utilize this matrix:

I will say a common error / problem to watch out for with this method though is race conditions. I had to toy with a decent notebook split to where GitHub actions didn't yell at me. It does take some tweaking, but we could speed up testing from 17 minutes to 7 :)

*For AdaptNLP