[Documentation] FX module is for Pass Writers

I am going through the documentation of recently released FX module to Pytorch release 1.10.

Can someone explain what does “pass writers” mean in this context?

FX is a toolkit for pass writers to facilitate Python-to-Python transformation of nn.Module instances.

I tried looking up the term “pass writers” with no relevant result that can tie it back to FX module.

Thanks!

It’s just meant to reference a person (a “writer”) who writes some transformation (a “pass”) for the FX IR. E.g. relevant wiki page on a compiler that does multiple passes:

1 Like

Appreciate your ELI5 explanation, @jfix! Thank you.