I’ve been having some problems completing this course. My main problem is that I really don’t want to use the Fast.ai library, but instead something like PyTorch or Keras. I feel like the Fast.ai library is way too high level, and I’d really like to learn some tool that I’m going to use afterward, like PyTorch or Keras.
I understand that the development of this library has taken much effort and that there most likely won’t be any official PyTorch/Keras notebook implementations anytime soon, but I’d like to know if you guys know of any additional resources, so that I could continue with the lessons?
There is a thread about implementing fast.ai in pure Pytorch: Plain PyTorch implementation of fast.ai notebooks
To learn Pytorch you can start by the Udacity course here: https://www.udacity.com/course/deep-learning-pytorch--ud188
It is a hard work reimplementing fast.ai but I think you will learn a lot when you finish. I have done one about Tabular module (v 0.7) Reverse Tabular Module in fast.ai v1 . with other module like Vision or Text it would be much harder
Hope that helps,
This certainly could be a great way to learn PyTorch, but I think I’ll need more hand holding before that. Have you completed that Facebook’s DL with PyTorch course? I couldn’t really find any reviews of it, so I was wondering what it’s actually like.
For me what worked best was to start playing with pytorch from very basic examples. You can start by implementing a linear regression model, there are plenty of examples online. It’s a good example to understand all the steps needed to build and train a model. Then you start also to understand fastai better.
Thanks! I’ll definitely give it a go!