should generate_quantities accept sample from Pathfinder as input?

[mitzimorris]

Summary:

generate_quantities has been generalized to accept samples from Pathfinder?

Description:

requires changes to cmdstanpy/model.py method generate_quantities:

cmdstanpy/cmdstanpy/model.py

Line 1272 in d4839e0

if isinstance(previous_fit, (CmdStanMCMC, CmdStanMLE, CmdStanVB)):

Additional Information:

add this as part of planned 1.2 release?

Current Version:

Please include the output of import cmdstanpy; cmdstanpy.show_versions(), or
at least the cmdstan and cmdstanpy versions used.

[mitzimorris]

@WardBrian : 👀 ?

[WardBrian]

It wouldn’t be that hard IMO but I’m also not willing to hold up the release of 1.2 for it. If someone asks for it we can do it

[mitzimorris]

OK, because looking at it now, it seems pretty easy to do - I could probably bang it out (famous last words...)

[mitzimorris]

same question for a laplace sample.

[WardBrian]

No technical reason either should be hard (in fact, they’re conceptually easier than the cases we already handle like ADVI, which has the mean as the first output).

It does still touch a lot of the code in the gq class (mostly updating if/else branches). I still think it would be nice to come up with a better encapsulation of standalone GQ

[mitzimorris]

agreed. let's wait on this

[WardBrian]

Sounds good, good to have an issue to track it.

If the docs updates are ready and you're able to review #691, I think we can release 1.2 early next week