fix: correct biased pymc HMC sampling (target_accept) - #1908
Merged
Conversation
The hmc parametrizations of test_c2st_pymc_sampler_on_Gaussian were failing in CD (c2st ~0.62). Root cause is PyMC's HamiltonianMC default target_accept=0.65, which mixes poorly on this peaked target regardless of seed, warmup, or number of draws (NUTS is unaffected). Expose seed and target_accept on PyMCSampler and set target_accept=0.95 (c2st ~0.51-0.54 across seeds and chain counts) plus a fixed seed in the test for reproducibility.
MCMCPosterior._pymc_mcmc left PyMC's HamiltonianMC at its default target_accept=0.65, which produces biased hmc_pymc posteriors (c2st ~0.62 on a Gaussian). Default it to 0.9 for the HMC step, leaving nuts_pymc and slice_pymc unchanged. hmc_pymc now samples accurately (c2st ~0.51).
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #1908 +/- ##
==========================================
+ Coverage 87.81% 87.94% +0.13%
==========================================
Files 143 143
Lines 13387 13657 +270
==========================================
+ Hits 11756 12011 +255
- Misses 1631 1646 +15
Flags with carried forward coverage won't be shown. Click here to find out more.
|
dgedon
reviewed
Jul 14, 2026
Contributor
Author
|
thanks for the review @dgedon when addressing your comments I noticed there was an inconsistency in the default values (0.9 vs 0.95 in the tests), that's fixed now. And I noticed it would be better to have the new |
dgedon
approved these changes
Jul 24, 2026
dgedon
left a comment
dgedon
left a comment
Collaborator
There was a problem hiding this comment.
Maybe update the original description of the PR, so it reflects the latest changes, e.g. the 0.9 default.
I approve for now since all remaining comments are smaller details and not functional really. In that way you can merge easier @janfb
Addresses review feedback from @dgedon on #1908: - add `validate_target_accept()` to `sbi/utils/typechecks.py` and use it at all four validation sites, replacing the repeated `validate_float_range` block. The helper lives in `typechecks.py` because it has no sbi-domain imports, so `MCMCPosterior` and `MCMCPosteriorParameters` can use it without pulling in the optional `pymc` dependency. - `MCMCPosterior.sample()` now validates only the caller's argument; the fallback to `self.target_accept` is already validated in `__init__`. - name the HMC default `_DEFAULT_HMC_TARGET_ACCEPT` in `pymc_wrapper.py` instead of hardcoding `0.9`, with the rationale on the constant. - state explicitly in all docstrings that `target_accept` is used only by `hmc_pymc` and `nuts_pymc`, and ignored by `slice_pymc`, the Pyro samplers and the numpy slice samplers. - revert unrelated whitespace and variable churn in the Pyro/PyMC branches.
Validate the type and (0, 1) range directly instead of delegating to validate_float_range. Error types and messages are unchanged.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follow-up to #1894 because CD is failing again. Once the lockfile fix let the pymc
tests run again in CD,
test_c2st_pymc_sampler_on_Gaussianfailed on thehmcparametrizations with c2st around 0.62.
However, it's not flaky: pymc's
HamiltonianMCdefaults totarget_accept=0.65,which mixes poorly on this peaked Gaussian and biases the samples, regardless of
seed, warmup, or number of draws. NUTS works fine.
To fix this:
seedandtarget_acceptonPyMCSamplerhmc_pymctotarget_accept=0.9(_DEFAULT_HMC_TARGET_ACCEPTinpymc_wrapper.py) to get accurate HMC posteriors (c2st ~0.51 vs ~0.62);nuts_pymckeeps PyMC's backend default andslice_pymcis unchangedMCMCPosterior(target_accept=...),MCMCPosterior.sample(target_accept=...)andMCMCPosteriorParameters.target_accept, validated to be in(0, 1)0.9default rather than passing its own value
target_acceptis PyMC-only:pymc.Slicedoes not take it, so it is forwardedonly for the
hmc/nutssteps, and the Pyro samplers (hmc_pyro,nuts_pyro)ignore it entirely.
Since this silently changes a default, there is a CHANGELOG entry under
v0.26.1→ Bug Fixes.Note: the c2st test is marked
slow, I will run the CD on this branch viaworkflow dispatch.