Compat for Turing v0.41 #44
Conversation
| vi = Turing.DynamicPPL.VarInfo(rng, model, Turing.SampleFromUniform()) | ||
| vi_spl = last(Turing.DynamicPPL.evaluate_and_sample!!(rng, model, vi, Turing.SampleFromUniform())) | ||
| θ = vi_spl[:] | ||
| ℓp = LogDensityProblems.logdensity(ℓ, θ) |
There was a problem hiding this comment.
calling logdensity leads to an extra model evaluation, we can skip this by using the logjoint which is already in the varinfo
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #44 +/- ##
==========================================
- Coverage 96.30% 96.29% -0.02%
==========================================
Files 10 10
Lines 352 351 -1
==========================================
- Hits 339 338 -1
Misses 13 13 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
SliceSampling.jl documentation for PR #44 is available at: |
| uses: TuringLang/actions/DocsDocumenter@main | ||
| with: | ||
| julia-version: 1.11 |
There was a problem hiding this comment.
the docs example uses PG, which doesn't work on 1.12 so this has to be pinned for now
There was a problem hiding this comment.
Hi @penelopeysm , thanks for the PR. Everything looks good. One quick thing though, could you update the docstring here, which seems outdated now that we have a more explicit way to initialize the parameters?
|
Thanks @Red-Portal! I just deleted it because initial params are now always in unlinked space so there's no possibility of the params being different from what the user specified. I think. Does slice sampling require linked VarInfo / unconstrained parameters? |
|
@penelopeysm thanks for the update. All looks good to me now.
Except for bounded supports, in general, yes. |
|
Neat, okay. There might be some upcoming changes to the external sampler interface, but it will be a while. I'll check back again when those land! |
No description provided.