Repository navigation
test(nuts): bound the divergence rate instead of requiring exactly zero - #215
Merged
Merged
Conversation
NUTS trajectories differ across platforms: the M.2F block in test-mixed-nuts.R has one divergent transition on the Linux x86 runner and none on arm64, with identical seeds, and the weekly certification job has failed on it every Sunday since 2026-08-09. The sampler-agreement comparisons in that block pass on both. The ten exact-zero assertions in test-mixed-nuts.R and test-ggm-nuts.R now require the divergence rate the diagnostics already compute to stay below 0.1% of post-warmup transitions.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
NUTS trajectories differ across platforms. The M.2F block in test-mixed-nuts.R has one divergent transition on the Linux x86 runner and none on arm64 with identical seeds, and the weekly certification job has failed on that single assertion every Sunday since 2026-08-09; the sampler-agreement comparisons in the same block pass on both platforms.
Change: the ten assertions in test-mixed-nuts.R and test-ggm-nuts.R that required
total_divergences == 0now requiredivergence_rate < 1e-3(the rate the diagnostics already report; 0.1% of post-warmup transitions). One header note per file states the rule. No package code changes.Verification, locally under the certification gate (BGMS_RUN_SLOW_TESTS and BGMS_RUN_CERTIFICATION set): test-mixed-nuts.R 254 pass / 0 fail, test-ggm-nuts.R 82 pass / 0 fail.
The other weekly failure, the difference-scale reweighting block in test-prior-sensitivity.R, is not touched here: it measures a real limitation of reweighting at a doubled scale and is addressed by the prior-sensitivity engine design.