Which version to use?

Hi all, I’m wondering if there’s any general guidelines for which version(s) of fastai correspond to which use case. There’s been a few posts lately where certain functions (draw_tree and plot_top_losses) were not working without some extra steps. I seem to remember that “fastai” was kind of a beta/developer thing - does that mean fastai-v4 is more ready out of the box? Would it still work alongside fastchan? Any clarification would be helpful. Thanks.

Generally, the fast.ai library v2 (current latest version) (GitHub - fastai/fastai: The fastai deep learning library) is what you should be using. It is the latest version and is being maintained. v1 of the fast.ai library is not being maintained (GitHub - fastai/fastai_old: OLD REPO - PLEASE USE fastai/fastai). fast.ai library v2 goes along with the 2020 (v4 i think?) version of the course which is based on (GitHub - fastai/fastbook: The fastai book, published as Jupyter Notebooks). There is still good information contained in previous courses prior to 2020 Part 1, but getting that code to run without some in-depth knowledge might be difficult. The 2019 Part 2 course is worth checking out if you’re looking for a much more in depth deep dive but it’s not based on the current version of the library so getting examples from it to run will be more challenging. Hopefully this helps.

1 Like