计划执行与检查节点
executing-plans
按既定实施计划完成工作，保持过程检查和反馈。
探索技能
效率与协作 · SKILL.md
mas_unlabeled_importance
mas_unlabeled_importance【效率与协作】Estimate Memory Aware Synapses parameter importance from unlabeled inputs using output-sensitivity gradients.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
技能说明
GitHubname: mas_unlabeled_importance description: Estimate Memory Aware Synapses parameter importance from unlabeled inputs using output-sensitivity gradients.
MAS Unlabeled Importance
Use when MAS recovery needs label-free importance. Inputs are weights and unlabeled samples; outputs are nonnegative importance values and sample-count metadata. Compute squared-output-norm gradients, average absolute values, and reject empty streams. Validate with python tests/test_importance.py.