fastbook VS nbdev

This chapter focuses on defining the DataLoader classes and a Bing Image Search downloader that is provided with the fastai library. We're not going to implement a Bing downloader. That is too much work for something that could be a crate on its own. Please feel free to write such a crate, though, the world could use one.

The book is available online here The course is accessible here

Fastai Lesson 2 notebook

I recently read the FastAI book from O'Reilly, which is also published as a series of notebooks on GitHub here. I personally liked it because it shows how to obtain a working model trained with modern techniques without delving too much in the low-level details.

I recommend fastai-fastbook. I just started myself though it’s a coupled with tools and a way of working that may help you including being and to create and publish python packages from a jupyter notebook using nbdev

The fastai book actually makes a nice comparison between the systems described in PDP and modern deep learning.

> In fact, the approach laid out in PDP is very similar to the approach used in today's neural networks.

From: https://github.com/fastai/fastbook/blob/master/01_intro.ipyn...

I’m a fan of fast book by fastai.

One example I've seen is ML/DL folks using jupyter notebooks to develop DL libraries in jupyter notebooks, see https://github.com/fastai/nbdev

- https://github.com/fastai/nbdev: Increase developer productivity by 10x with a new exploratory programming workflow.

9. Nbdev: Boost developer productivity with an exploratory programming workflow - https://nbdev.fast.ai/

See if you like nbdev way of working with data through python and jupyter. nbdev is an optional part that will create python packages from jupyter notebooks. Also even the simple tutorials are opinionated and will guide you to unit test your code and write CICD pipelines.

You probably don’t have time for this for the purposes of your task, but I will also throw in the recommendation of nbdev especially if you’re a Python person. I haven’t had a project to use it on yet, but I’ve gone through the docs and the walkthrough and it seems like a great framework for starting potential projects with all the infrastructure needed for if/when they eventually get big and need all the packaging and stuff

nbdev: jupyter notebooks -> python package

Take a look at https://github.com/fastai/nbdev - haven't used it but supposedly the whole if fast.ai library was written that way. It sounds like a natural direction in your scenario - allowing your to keep working in a familiar environment and still producing production ready code (will, at least in paper 😅)