ponder-lab / Hybridize-Functions-Refactoring / #2389

81%

main: 83%

Build Type

Pull #750

github

Committed by khatchad

Commit Message

Add value-slice regression test, gated-check pin, and named constants

- testNumpyOnSlicedTensorValue pins the RPN compute_iou regression: numpy over
  a slice of a tensor value (`np.maximum(boxes[:2], ...)`) stays a sink.
- Re-add the einsum_via_matmul assertion, reversed: its numpy check is gated
  off by its primitive parameter, so getHasNumpyCallsOnParameters() is null.
- Extract the "slice" builtin name and the "neg" operator name to named
  constants.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01He1ULx8nRoHzGicMryry1A

Pull Request Pull Request #750: Make the parameter-flow numpy precondition per-dimension shape-aware

Coverage Stats

171 of 196 new or added lines in 3 files covered. (87.24%)

59 existing lines in 2 files now uncovered.

2061 of 2546 relevant lines covered (80.95%)

0.81 hits per line