Conversation
Introduced a new `MicroPoints` data structure and incorporated micro-point corrections into the interpolation framework. This includes gradient computation, anisotropy handling, and validation for authored points, along with warnings for unsupported or ignored configurations.
This was referenced Sep 28, 2026
Merged
Member
Author
Refactored micro-point correction into a separate `MicroCorrection` class for modularity and maintainability. Replaced the earlier in-place configuration logic with stack-local `MicroCorrection` objects. Updated references, tests, and removed legacy micro-point settings. Improved evaluation performance by optimizing gradient computation and kernel handling.
Modularized micro-point handling by introducing stack-local configuration via `prepare_micro_points`. Extended support for authored micro-points in fault and non-fault stacks, adding validation, residual computation, and warnings for unsupported configurations. Enhanced evaluation flow by integrating grid, macro, and micro point structures, optimizing kernel compositions, and ensuring consistent coordinate layout across interpolation workflows. Updated documentation and adapted tests for new behaviors and configurations.
- Fixed NumPy-authored micro coordinates entering PyTorch evaluation, including CUDA and autograd. - Removed the test-only correction hook from _evaluate_sys_eq(). Macro evaluators no longer depend on micro correction. - Combined correction values and optional gradients into one evaluation pass. - Centralized global ownership validation and unsupported-stack checks in preparation, removing duplicate checks from stack execution. - Documented global versus stack-local surface indices and prepared query state. - Moved prototype solve helpers into a test-only reference module and removed the unused macro-evaluation helper. - Added transform, backend, early-validation, and multi-level mesh regression tests.
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.
Introduced a new
MicroPointsdata structure and incorporated micro-point corrections into the interpolation framework. This includes gradient computation, anisotropy handling, and validation for authored points, along with warnings for unsupported or ignored configurations.