-
Notifications
You must be signed in to change notification settings - Fork 104
Add page on models and varinfos, part 1 #663
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
3ade052 to
1e3f53f
Compare
|
Preview the changes: https://turinglang.org/docs/pr-previews/663 |
mhauru
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A tiny optional comma-move, otherwise I'm happy to merge. Maybe hold off for a few days in case someone else wants to comment.
I have many things I would like to add to this, but I think it makes sense to expand later once the current VarInfo overhaul is done.
Feel free to tack on to #664 / a new issue! |
Co-authored-by: Markus Hauru <mhauru@turing.ac.uk>
|
Merging, any other suggestions from anyone, please feel free to open a new PR |
Closes #660.
Preview @ https://turinglang.org/docs/pr-previews/663/developers/models/varinfo-overview/
I tried very hard to avoid going into untyped vs typed VarInfo, but it proved to be impossible to fully explain this without mentioning that, because some things simply can't be done with typed VarInfo e.g. TuringLang/DynamicPPL.jl#1062.
However, it is self-contained at the very end of the page and for the most part, we only cover four things:
VarInfo(model)evaluate!!init!!unflattenwhich mostly lines up with my plans from #660.
I didn't make any mention of
Metadatathe struct, so even if we replace Metadata with VNV, the considerations described here should still hold.This is not future-proof against a complete overhaul of VarInfo, but I assume that will take a while and in my opinion it's a Good Thing to have some docs in place now.