Thank you all SO much for joining me tonight, it was a ton of fun to be back in the streaming/teaching wheelhouse again! We had almost 300 people chime in! For those that missed out, the video link is here:
1/
The notebooks are also available here, and eventually I'll have them fully integrated into walkwithfastai.com, just need to figure out a good way first! 2/ github.com/walkwithfastai…
From what I gathered from the feedback, it seems this stream really did help open up everyone's eyes to the power of @fastdotai's Callback system, and how it's training loop works. It's not just a marketing phrase when they say you can modify ~everything~ in the training loop! 3/
Again thank you all so much, it was a bit different than what I was used to (despite the course, little to none of it was direct live coding like we did), so thank you for bearing with me! 4/
Next week we're going to hit @fastdotai loss functions, what magic is inside it, and why suddenly people are getting these very odd "X is not implemented for Y" errors. See you then!
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Have you wanted to know just what goes on in @fastdotai's training loop and just what those pesky Callbacks are up to? In the latest release of Walk with @fastdotai this is now possible!
What we've done here is extended fastai's existing `show_training_loop` functionality to include some information (from a provided doc string) of what exactly is occurring during each event. This then lets you get more familiar with what fastai's training loop, ... 2/
Understand when things are called, and what happens at each step. The old version (which is still there if you pass `verbose=False`) simply just showed what Callbacks were triggered (pictured below)
At the end of the day, it's not about who copied who. It's about taking and using the ideas of some other library (which also is published paper), without giving the proper attribution in your GitHub or documentation, which gets the most traffic. 1/7
And don't even *try* to claim to do something revolutionary, when in reality it's something others have taken months and years to build, and you’re simply taking notes (or more). 2/7
Open source frameworks and the evolution of these libraries build upon each other. That’s the definition of a community. You look at what’s being used, where their weakness are, their strengths, and you start tweaking. 3/7
Today I’ve come across some wonderful @fastdotai YouTube channels with some excellent content! Below is a thread of my findings for folks to check out (I’ve subscribed to them all!) 1/
First and foremost we have the wonderful work coming out of @ai_fast_track. Along with the #IceVision videos he’s also done quite a number of videos exploring the @fastdotai API with some EXCELLENT videos, I’m certainly taking notes youtube.com/channel/UCht9j…
Next we have some videos by @philwhln. His first two short videos on #fastbook show some great insights into dealing with issues he had, and a great overview of the first few chapters 2/ youtube.com/user/philwhln
The first introduces the library and quickly examines what's new, a perfect start for beginners and experts alike. We will cover just what @fastdotai is, how its used, and the goal of the library: muellerzr.github.io/fastblog/2020/… 2/6
The second assumes that you are familair with the original library (@fastdotai v1). We'll compare 1:1 examples of the API, go deep into the High-Level API, and discuss what some of the best new tricks are: muellerzr.github.io/fastblog/2020/… 3/6