Compute Metropolis average accept probability with logsumexp
#7783
+4
−2
Conversation
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
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #7783 +/- ##
==========================================
+ Coverage 92.83% 92.84% +0.01%
==========================================
Files 107 107
Lines 18354 18381 +27
==========================================
+ Hits 17039 17066 +27
Misses 1315 1315
🚀 New features to boost your workflow:
|
ricardoV94
reviewed
May 21, 2025
ricardoV94
approved these changes
May 21, 2025
|
@ricardoV94 any thoughts about just clipping the acceptance probability to be in 0,1 ? |
Co-authored-by: Ricardo Vieira <28983449+ricardoV94@users.noreply.github.com>
Is that why you were getting nans after the exp? Shouldn't hurt, as long as it's all cheap stuff |
|
Yes I assume it was overflowing. In the progress bar I was seeing large values for acceptance |
Then feel free to clip the exp from above or the log to be at most 0 for the returned stat |
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Outdated suggestions cannot be applied.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.
In the Metropolis step stats that we report on the progressbar, we have a probability of acceptance. This is currently computed in a very direct way, just
np.mean(np.exp(log_accept_prob)). We can do better withnp.exp(logsumexp(log_accept_prob) - log(len(log_accept_prob))).In really gnarly models, I've encountered overflows in this step. This might also be because there's a max(delta_logp, 1) missing from the acceptance probabilities. But even if we add that, the computation method in this PR is more numerically stable and accurate.
📚 Documentation preview 📚: https://pymc--7783.org.readthedocs.build/en/7783/