Fix uninitialized norm bug and bump version to 0.10.1 - #10
Merged
Conversation
Rebuilt against the committed norm(numeric_limits<double>::infinity()) fix (62495bb) and ran every check in test1.py/test1_pip.py, extracting actual values instead of relying on pass/fail alone (test files were temporarily instrumented for this, not committed). Only 4 of 84 checks actually changed - all four are the average-reward MPI/PI (parallel and unparallel) value-vector checks in Configuration 1 (Model 1a/2a/1b/2b). Every other check, including the discounted- criterion ones, still passes: differences there are ~1e-4 and round to the same displayed value, well within existing tolerance. The four new values are each the old value plus a constant ~13.4623 across all three states - exactly the expected signature of shifting the average-reward differential value function's additive-constant anchor (see the norm-uninitialized-bug-deferred memory note). The policy in every case is unchanged: [1, 1, 0]. test2.py/test2_pip.py need no changes - they only use the discounted criterion, which doesn't have this ambiguity. Verified: all of test1.py, test1_pip.py, test2.py, test2_pip.py pass against a fresh local build with both this change and the value-vector updates applied. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Patch bump: fixes the uninitialized norm bug in ModifiedPolicyIteration (62495bb) and regenerates the affected average-reward test oracle values (f84c13e). No new features or API changes. Verified via a manual workflow_dispatch run: build+test passes on Linux, Windows, and macOS. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
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
1 participant
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.
Fixes a real undefined-behavior bug (
normread before initialization inModifiedPolicyIteration, found via Valgrind) and regenerates the average-reward test reference values it affected. Verified via a manual workflow_dispatch run across Linux, Windows, and macOS before merging. Bumps version to 0.10.1 (patch-bug fix only).