Fastai documentation \o///

Thought I’d start a topic to talk about documentation for the fastai library in general, since we don’t have any other so far.

Being honest though, my hidden goal is to thank everybody who has been involved in writing/updating the documenation, and making it available online. This is just amazing, as it’s going to help immensely understanding the library design. Thanks again :clap:

Looking forward to the lectures, and contributing to the documentation where needed.

Links for documentation:

Also, I just remembered that Jeremy shared this ‘Introduction to PyTorch’ notebook just a while ago. Adding it here as a bonus/related link.


Your appreciation is much appreciated. :slight_smile: We’re all really proud of it, and of the new docs framework we used to write it.

There’s still plenty of room for improvement of course, so contributions are most welcome!


Thanks for the great work guys!

Love the fact that it’s generated using Jupyter notebooks that we can clone and run, to follow along as we read. Can’t believe I’m actually looking forward to reading a library’s docs!


Maybe I’m the only one who doesn’t know that there’s a documentation for developers.

Why is not working? @jeremy @sgugger

Not our fault:
It should be back sometime soon.


Once again, maybe it’s just me, but the docs are amazing. Perfect for not having to jump in the source or the forum right away when there’s some confusion. Also, the explanantion helps a lot.

Now and then, I run into links that are broken, but I’m assuming this is because some part of the libs are in flux, or should I be reporting this somewhere ? (I don’t want to create more noise either)

Please do let us know - on #fastai-users:fastai-dev .

Just a minor enhancement, i do see some help doesn’t show the return value or type ; So i have to take a look at the source code ( which is a good thing ).

[1] help(untar_data) <-- No return value in the help o/p
Help on function untar_data in module fastai.datasets:

untar_data(url:str, fname:Union[pathlib.Path, str]=None, dest:Union[pathlib.Path, str]=None, data=True)
Download url if doesn’t exist to fname and un-tgz to folder dest

[2] help(get_transforms) <-- this shows return value in help o/p

Help on function get_transforms in module

get_transforms(do_flip:bool=True, flip_vert:bool=False, max_rotate:float=10.0, max_zoom:float=1.1, max_lighting:float=0.2, max_warp:float=0.2, p_affine:float=0.75, p_lighting:float=0.75, xtra_tfms:float=None) -> Collection[]
Utility func to easily create a list of flip, rotate, zoom, warp, lighting transforms.

Use doc instead of help.

Shouldn’t this section( include
gradient clipping ( too ?

I think so! Would you be interested in trying to do a PR to fix that?

Sure. In a minute :slight_smile:

getting error trying installing the doc tools process following this:

(base) bherudek-ltm1:fastai_docs bherudek$ tools/run-after-git-clone
File “tools/run-after-git-clone”, line 18
cmd = f"{sys.executable} {script}"
SyntaxError: invalid syntax

Any ideas?

What python version you are using ? Make sure you are using >= 3.6