Cool! If you find any particularly compelling models, let me know so we can try them in Stan too.
Awesome! Looking forward to teamworking this. And the delay is alright. I’ll be at work and good chance I don’t get anything done anyway :P.
The pull request right now is exclusively switching between diagonal and dense metrics as defined in Stan. I figured I’d just go conservative on what we try to get in. But I think it’s fair to be fancy about how we pick between those two (esp. if it makes things more robust).