ewc_recovery_evaluation
Run a bounded EWC retention recovery comparison and emit validator-compatible result, trace, and mechanism-check artifacts.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- ewc_recovery_evaluation
- description
- Run a bounded EWC retention recovery comparison and emit validator-compatible result, trace, and mechanism-check artifacts.
02 资源文件 2
- 脚本scripts/recovery_evaluation.py9.4 KB
- 脚本tests/test_recovery_evaluation.py1.1 KB
03 SKILL.md 正文
EWC Recovery Evaluation
Use this skill to execute a reduced or full sequential-training comparison for Elastic Weight Consolidation. It should be used after task protocol, Fisher importance, and EWC penalty skills are available.
Inputs
- Protocol JSON with task order and datasets.
- Generated skill root containing
ewc_task_protocol,ewc_fisher_importance, andewc_penalty. - Module-plan target metadata.
- Output paths for recovery artifacts.
Outputs
recovery_result.jsonwith numeric retention metrics and mechanism checks.training_trace.jsonwith before/after loss and parameter values.- Skill invocation evidence proving generated modules were used.
Workflow
- Fit task A and snapshot parameters.
- Estimate a diagonal Fisher from task-A gradients.
- Train task B with plain SGD and with EWC.
- Compare task-A retention and Fisher-weighted drift.
- Save command-produced recovery artifacts.
Validation
Run:
python tests/test_recovery_evaluation.py
Limitations
The bundled harness is a reduced synthetic proxy. It does not claim full permuted-MNIST or Atari reproduction.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).