@hiimnick and I are going through the fast.ai course and building projects based on what we’re learning in the course. On Mondays we watch and discuss lectures and on Thursdays we build projects based on the previous lecture.
This is not a traditional study group. Everyone is welcome to drop by as much as they want to and participate through our twitch stream. We’re both programmers, but we both don’t know fast.ai yet, so expect debugging, detours, and projects. ¯\_(ツ)_/¯
Watch: Mondays, US 12:00PM EST / EU 17:00PM CEST: Watching, reading, and discussing the lessons
Build: Fridays, US 12:00PM EST / EU 17:00PM CEST: Building projects based on learned concepts
If you missed out on one of the streams, you can re-watch them on our YouTube account:
Upvote or spread the word if you like this idea, the bigger our twitch stream learning group gets, the more fun we’ll have together
Thanks for everyone who joined in yesterday (and for letting us know that my audio was off ) We’re putting timestamps on our previous streams right now and gonna have them up on YouTube in a few days.
Looking forward to building out our first applications from the pickled models on Thursday! Hope to see you around : )
We also managed to upload a couple of our previous streams to the YouTube channel and added nice timestamps with emojis to make it easier to skip to the parts that might interest you. Check it out:
Today we’re going to continue to read over Lesson 3, starting at SGD, so we’re still in the notebook named 04_mnist_basics.ipynb. We’re going to keep reading the lessons together as we did in our last stream, and consult the video only if we want some additional clarification. Seemed like more fun like this
Next livestream starts in 30 minutes! We’ll learn more about SGD by hacking the scraps of theoretical knowledge we gained last time onto the MNIST Loss Function, and probably do a little recap of SGD on the way. Here’s a link to the relevant section in the notebook: 04_mnist_basics.ipynb (you can link to headers in a notebook, wow! )
The next livestream starts in 30 minutes! Today we’re gonna hop back into the first notebook 01_intro.ipynb and attempt to figure out the differences between:
fine_tune()
fit_one_cycle()
fit()
By just running all of them as often as possible and making up weird theories of why things happen. We’re probably also gonna look into the docs, though. Join us at 11AM EST / 5PM CET!
30 minutes to our next live stream! Today we go over Chapter 05_pet_breeds and see what of the content we can apply to our own dog breed classifier that we built back in Episode 2.
30 minutes to our next live stream! Today we keep working on Chapter 05_pet_breeds and continue to apply the content to our ancient history dog breed classifier that we built back in Episode 2.
After our Thanksgiving stream disaster, join us trying to learn some more anyways. Getting better is incremental and takes time and tenacity is key, right? ¯\_(ツ)_/¯
Hop on our next live stream in 30 minutes! Today we keep moving ahead with the content going through the first part of Chapter 06 - Multi-Category.
Hop on our next live learning stream in 45 minutes! Today we keep moving ahead with the content going through the second part of Chapter 06 - Multi-Category, which is about Regression.
Detour Alert Today we will take a detour and explore Kaggle. Both @hiimnick and I know about Kaggle for a long time, but never really sat down to explore it in more depth. Today’s the day.
We’ll look at the Planet dataset and figure out whether that is a project we want to tackle coming up next.
More on the Planet dataset . @hiimnick downloaded the data to our GDrive (we’ve got 3GB space left ) so we’ll spend today exploring the data using fast.ai and whatever again was that elusive code to display images and get stats using pandas.
No final decisions yet, but so far it looks like a fun and interesting project to learn more about!
Still looking at our Planet (dataset) . Turns out that @hiimnick downloaded something last time, but the data wasn’t in a form we could use. So maybe we got it today… ¯\_(ツ)_/¯
Also, if you just want a small version of the dataset, you can just use… fast.ai:
The holidays were long, we completely forgot what we were doing last time. So this stream we’ll get back on the saddle and figure out what we did and what we’ll do moving forward:
Join us if you want to! Everyone’s welcome in 2021!
Join our fast.ai learning stream on twitch.tv:
) We’re putting timestamps on our previous streams right now and gonna have them up on YouTube in a few days.
We’ll post a link when we get it done : )
to make it easier to skip to the parts that might interest you. Check it out:
)
We’re probably also gonna look into the docs, though.
!
!
classifier:
WE’RE BACK! 
, join us trying to learn some more anyways. Getting better is incremental and takes time and tenacity is key, right? ¯\_(ツ)_/¯
Detour Alert 
) so we’ll spend today exploring the data using fast.ai and whatever again was that elusive code to display images and get stats using 
