Skip to content

V0.3.1 unify gain computation#9

Merged
omereliy merged 3 commits into
masterfrom
v0.3.1-unify-gain-computation
Mar 22, 2026
Merged

V0.3.1 unify gain computation#9
omereliy merged 3 commits into
masterfrom
v0.3.1-unify-gain-computation

Conversation

@omereliy

Copy link
Copy Markdown
Owner

No description provided.

Sequential and parallel gain paths now use the same
_compute_single_action_gain function via SequentialGainCache,
fixing a lifted-vs-grounded divergence in success potential.
Export/PDDL/metrics code moved to model_export.py, reducing
information_gain.py from 2200 to 1823 lines.

- Fix _calculate_potential_gain_success to use grounded-space
- Add SequentialGainCache for in-process gain delegation
- Add _create_sequential_context (references, no deep copies)
- Deterministic tie-breaking in both sort paths
- Extract to_pddl_string, export_model_snapshot, get_learned_model,
  get_action_model_metrics to algorithms/model_export.py
- Bump version to 0.3.1
- Delete pddl_types.py, inline ParameterBinding into expression_converter.py
  (one-consumer rule: only expression_converter used it)
- Remove unused ExpressionConverter.to_cnf_clauses() and its tests
  (no production callers)
- Update imports in up_adapter.py and test files
Combine model_validator.py + model_metrics.py + model_reconstructor.py
into a single model_evaluation.py (-420 LOC net). These are tightly
coupled BRIDGE modules always used together. Breaking old import paths
is fine at v0.x.
@omereliy omereliy merged commit 6d183b4 into master Mar 22, 2026
3 checks passed
@omereliy omereliy deleted the v0.3.1-unify-gain-computation branch April 3, 2026 14:05
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant