hey… i am a beginner to both pytorch as well as fastai. I know certain basic theories in ml and dl but i wish to get competent at coding with these as well. So i have just completed lesson4 of fastai v3 part1.
Can you please what could be a good learning path.
I was thinking to start the 2018 ml fastai course to get better at ml and v3 for deep learning. Should i continue with the same or would you suggest something else?
Hi @Adarsh97
I wouldn’t recommend the 2018 Machine Learning course, its way too outdated now. You can continue with the 2019 deep learning course. That being said, keep doing course v3 and, and simply start from the beginning when course v4 comes out. You’ll only learn more with The v4 course combines the Machine Learning course from 2018, and Deep Learning. It should be out in the next few weeks.
oh ok thanks! i just hope the v3 course that i am doing right now doesn’t go to waste once fastai 2.0 comes out.
Also, i am thinking to learn pytorch as well. Can you please recommend some good resources for the same?
It won’t go waste! No effort is ever waste.
About PyTorch…I don’t understand what you mean by ‘learning’ PyTorch. If you’re talking about getting familiar with all functionalities present in PyTorch, you might grab the free pdf of the new book by PyTorch, “Deep Learning with PyTorch”. Though, I personally didnt find it very helpful. I just got comfortable with PyTorch through the fastai courses themselves - by looking up the documentations as and when i needed anything specific from PyTorch. So you might just follow the fastai course and expect to get familiar with PyTorch on the way! Cheers!
well… i feel like i should know enough pytorch to build basic neural networks and execute certain basic functionalities alongside. I feel things arent as straightforward as keras when it comes to building models and stuff. So i feel like a good understanding of pytorch would help me to understand the underlying implementations of fastai models as well write my own custom functionalities. I hope you get my point and if you do, do you still think fastai courses would be enough or are you aware of any good resources for what i just suggested?
You should try the second part of the Deep Learning course. It’s exactly what you’re looking for, and much more! But i’d first recommend (very strongly) that you finish and master the first part of the course. The second part is basically building fastai from scratch. And whats better than learning to build the fastai library, given its one the best Deep Learning libraries available?!
Though, its quite an intense course. If you’re only looking for a basic knowledge about how to build basic models in PyTorch, check out their book(if you’re a book guy) or their video demos (if you’re a video guy)
I am reading the fastbook right now (I am more of a book guy than videos) and I am really enjoying the book. I just completed reading and running first two chapters on colab.
I am currently following zach’s walk with fastai2 and also reading “Deep learning with Pytorch” book. I am actually more comfortable with pytorch and now trying to use fastai.