FastAI course v4 MOOC

Hi, out of curiosity, is the v4 of the MOOC course still planned to be released in mid-July?

1 Like

Hi! I am tryin gto use v4 with NLP on paperspace gradient (It is arleady released in paperspace VM’s)
I always get the
name ‘TextLMDataBunch’ is not defined
I actually want to use the fastai to classify my senteces from csv file into 6 categories. Dunno what to do with 4th version. Docs are obsolete. Could you give link on smth for v4

Yes.

I am lucky enough to be able to follow the Part1 V4 course live, and Jeremy confirmed in the first lesson that it will be released publicly in July, along with the official release of the book.

1 Like

Will the book cover more than the course, such as transformer architecture and use-cases, text generation, FastAiv2 building from scratch, low-level stuff and more?

1 Like

I don’t think the final version of the book will cover more than what is currently in the draft (which is available at https://github.com/fastai/fastbook)

I hope it is not going to happen as you said. What is currently in the draft is really an introduction to the introduction of deep learning.

The book is basically done, as it’ll be released here in a few months (June/July depending according to them), so they’ve already probably had to push everything. It is meant to be an introduction book so don’t expect anything too crazy, also the course follows the book as well, so there won’t be any surprise material that shows that you weren’t expecting. For part 1 at least if it’s not in the book, it wasn’t in the course. (Also we didn’t go over the entire book during it, as there’s far too much material for that)

I believe one core objective of the course is to get deeplearning in the hands of as many people as possible. That’s the priority and it means they have to start from the begining.

I think they did an awsome job with the book and I am really thankfull to the team.

And I don’t agree with you, the low-level code from scratch you are talking about is there in the book.

Many of us wish Jeremy would do a part 3, part 4 etc., but honnestly after you master all the concepts taught in the book and the MOOC, I am pretty confident you can get going on your own to some more advanced techniques, and also contribute to side projects with others on the forum.

3 Likes

Yes. The recordings have just finished and will be made available end of June

Hey everyone,

I just started fast.ai v3 today, but was wondering if it would be better to just wait until the v4 MOOC was released in June. I don’t have much experience in deep learning to be able to determine if it worth the wait. Are fastaiv2 and Fast.ai v4 significantly different than their predecessors? If so, is it best that I wait? I would appreciate any feedback.

Thanks in advance.

1 Like

In my opinion, taking the course v3 would not be a waste of your time. The difference between the course v3 part1 and the course v4 part1 are mainly about fastai software improvements (more robust internal code and the API is more harmonized), not about a complete change of API or philosophy.

I always find it very useful to learn something new in a few different ways, it helps reinforce the concepts. Therefore, I think it will help you if you can contrast Jeremy’s explainations from v3 to v4, which are slightly different but fondamentally the same.

Plus, you will get to learn from v3 part2 which is a masterpiece.

2 Likes

My recommendation would be to read the fastai book instead of doing v3. Start reading and learn as much as you can, then follow the MOOC this summer.

1 Like

I will suggest to complete course v3. As Antoine said above, it is truly a masterpiece. Even if v4 is a complete redesign, the core concepts won’t change much.

1 Like

Does anyone have non-fake news about dates for course2-v4, or about when course1-v4 will be released to the general public that they can share?

I have been watching Jeremy’s live coding sessions released last September.
So, I was wondering how different is fastai v4 MOOC from these wonderful videos.

Any idea?

1 Like