nutpie icon indicating copy to clipboard operation
nutpie copied to clipboard

Published 20 hours ago • verified publisher icon pymc-devs

Add log-likelhood calculation to sampling

Open fonnesbeck opened this issue 4 months ago • 4 comments

This allows nutpie to populate the log_likelihood group in ArviZ InferenceData objects when compute_log_likelihood=True is passed to compile_pymc_model().

Closes #150

fonnesbeck avatar Jul 18 '25 18:07 fonnesbeck

Works in numba mode, but not jax.

fonnesbeck avatar Jul 21 '25 22:07 fonnesbeck

Why not just use pm.compute_log_likelihood?

ricardoV94 avatar Jul 24 '25 16:07 ricardoV94

@ricardoV94 for calculating likelihoods during sampling? I thought that only worked after posterior sampling. I want to calculate them on the fly if I need them, rather than having another manual step.

fonnesbeck avatar Jul 25 '25 00:07 fonnesbeck

Add it as a model deterministic then?

ricardoV94 avatar Jul 25 '25 06:07 ricardoV94