documentation and training recipes

[zhhao1]

Thank you so much for the great work. I wonder if there are documentation and training examples for quick use.

[alvations]

+1 on the training recipes doc!

Esp. if there's any additional preprocessing steps to convert textfiles to fairseq2 trainable format like in fairseq 1.

[LPX55]

@alvations @zhhao1

Unless I'm utterly lost or have missed something completely, I'm pretty sure the docs that are linked from the repo is not up-to-date. Your best bet is probably to build them locally -- you'll find that there are references and tutorials that are much more helpful.

edit: Quick look at #870 has me thinking that they are aware and a fix is OTW. Didn't look too deep into the actual commits, so it could just be maintenance and updates for the local docs 🤷‍♂️

edit2: Static build @ https://github.com/facebookresearch/fairseq2/actions/runs/12172534450/job/33951385408?pr=870#step:14:22

[zyaoj]

@LPX55

thanks for the mention! Documentation is OTW and you can find the latest build with our nightly documentation.

@zhhao1 @alvations

Please take a look at this tutorial and let us know if it is helpful. We will mirror the stable version once the current set of documentation get consolidated. Thanks!

[zyaoj]

Closing this now. Please let us know if there's any documentation you want to see.