tilelang2ascend-case-simplifier

What problem does it solve? Operator test suites often contain too many input cases, slowing down validation cycles. This Skill trims the cases stored in an operator's .json file to at most ten representative entries without losing dtype, attribute, shape, or broadcasting coverage. ## Core Features & Use Cases - Automatic Target Detection: Locates the correct .json file by parsing the json_path reference inside model.py's get_input_groups() function, falling back to scanning the output directory. - Coverage-Preserving Reduction: Selects cases that collectively cover every dtype, attribute value category, tensor dimensionality, extreme shape size, and broadcasting pattern found in the original set. - Safe Backup Workflow: Creates a .json.bak backup before rewriting, enabling full-case restoration for later verification. - Use Case: After generating an Ascend C operator project with hundreds of test cases, run this Skill to shrink the suite to ten cases for fast iteration, then restore the backup for final full validation. ## Quick Start Simplify the test cases in my operator output directory so the JSON file keeps at most ten representative cases with full dtype and shape coverage.