In the real dev world you should be working out of docker images or something similar on a system that is suported (windows is not one). Most folks run Linux + Anaconda which allows for versioning of different packages. (this is what Jeremy meant from a dev perspective). From a course perspective, any of the platforms mentioned can also double as a development platform too. I use colab regularly for this purpose mixed in w/ nbdev. And when doing GPU things, this is 100% mandatory (IE either you use someone else’s host or you host your own server to some degree, WSL2 is trying to fix that but it’s not far enough yet)
1 Like
Thanks for clarifying.
Besides Colab, I have used fastai in my Linux machine using virtual environments. Had a seamless experience. (Haven’t tried for CV or NLP, though.)
I, too, have used Colab for development using fastai .
I have always used Linux for any kind of development, not just Deep Learning or fastai .
1 Like
Well, I used conda to create the environment for fastai 0.7.0 ML course on WSL 2 Ubuntu and it worked fine. Also, you don’t require Windows 10 insider for WSL 2 anymore. From Windows 10 version 2004 you can download WSL 2.
Some people have reported problems with the CUDA drivers not recognizing the GPU(GTX 1050) and garbage collection. I’ve yet to experience that with RTX 2060.
2 Likes
Hi @kogam22
Maybe a blog post or even a forum post here documenting what you did to get it to run might be helpful for those folks who want to run fastai 0.7 on a local jupyter environment.
A local jupyter environment is really convenient sometimes, especially when your internet connection is flaky.
Best regards,
Butch
2 Likes
Wonderful news!
Just received my RAM upgrade yesterday and I’m ready for the course.
Thanks everyone for such hard and wonderful work.
Although I started working through v3 a few weeks ago, decided to wait for v4 and I’m very much looking forward to August 21!
Cheers,
Giga
1 Like
Just wrote down a post. Feel free to edit it if something does not work.
3 Likes
Thanks @kogam22. You might want to add a note that CUDA isn’t supported in WSL2 without using an Insiders build and special drivers.
5 Likes
Yep. I quoted you. Thank you for the amazing things you’ve done.
Original:
I think I hope fastai will have is to teach people how to build model in pure pytorch rather than using its framework. I understand it has high accuracy and has all the tool. But I still hope that I can know how to build the model rather than living in the fastai framework.
For example, there many new state of art model that is very similar to many fastai framework that I can add to improve the result.
Edited:
I hope to know how to build complex model and fastai. I didn’t mean to remove Fastai from the course.
1 Like
It…does though? MNIST is from scratch, so is retinanet from the older course. And the resnet example too. And LSTM. Fastai is in PyTorch. So I’m confused by this. You’re also asking to remove the framework from a course on how to utilize the framework. Go take a PyTorch course if you are not pleased with the part 2’s and other courses and resources available.
6 Likes
@JonathanSum
I think you’ve not checked out Part 2 of the course. Its exactly what you’re looking for.
The part1 of the course is meant for getting familiar with the framework. The motivation behind that is to let everyone- even absolute beginners, build their own models. A lot of people lose motivation when they’re overloaded with concepts over concepts, on top of coding bugs that they’re not yet comfortable enough with to be able to solve them.
Part 2 of the course is meant to teach people how to build stuff from scratch, just as you are expecting. This course essentially follows a top-down approach, that is getting familiar with Deep Learning first, and then learning all the innards.
The point of fastai, the way I look at it, is not to teach how to use their own framework and make the learners handicapped, but to let **everyone ** experience Deep Learning first hand.
Most importantly, fastai is definitely not meant to be an exhaustive learning source. Think of it as a source that makes you ready for more complex things in the domain.
I hope that gives you a better insight.
6 Likes
@ PalaashAgrawal
I agree with your point on fastai. Fastai is making neural nets uncool again. For anyone who has not studied Deeplearning before, they can even do pixel-level classification and more in just one to three days with fastai. It is about making everyone able to use every power of deep learning.
2 Likes
I’ve completed 4 lessons in the v3 course, should I wait for v4, or complete v3 first?
I’m thinking I’ll complete v3, both parts, and then v4.
What would you guys suggest?
Overall, the concepts will be the same in v4 and v3-part1. Only the implementation will change. So, your approach seems good enough.
Super excited! Its coming soon!
2 Likes
Hi Zachary, I can’t find the updated v3 notebooks. Did they get lost when fastai2 was renamed to fastai? I’m searching for the superres without GAN notebook. If that’s not in the coming v4 course, it would be very useful for me to have an updated version.
Thank you!
They’re now under dev_nbs/course
2 Likes
Thank you 
Just waiting for the course to release and not gonna sleep until i complete first unit… super excited
9 Likes
Few hours until the launch of v4!!
I was very excited as it’s already 21st here, later realized it’s still almost midnight of 20th in the US.
Can’t wait! Thank you for putting all the hard work you guys did and your wonderful team!!
@rachel @jeremy @muellerzr
10 Likes