sbi_recovery_diagnostics
Validate reduced or full SBI recovery artifacts, source boundaries, mechanism checks, and metric reporting.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- sbi_recovery_diagnostics
- description
- Validate reduced or full SBI recovery artifacts, source boundaries, mechanism checks, and metric reporting.
02 资源文件 2
03 SKILL.md 正文
SBI Recovery Diagnostics
Use this skill when a Paper2Skills recovery for the sbi toolkit needs to declare whether it is full or reduced, prove that the original repository was not used, and validate that the simulator-training-posterior mechanism was actually exercised.
Do not use this skill as a replacement for running the recovery experiment. It checks artifacts produced by executable recovery commands.
Inputs
- Runtime handoff with full-runtime readiness and blockers.
- Recovery result with metrics, target metadata, and mechanism checks.
- Training trace and generated-data item for reduced recovery.
- Source manifest and generated-skill invocation logs.
Outputs
- Diagnostic JSON with
ok,errors, andwarnings. - A concise interpretation of whether the recovery scope is valid for soft or hard mode.
Workflow
- Check that full-runtime claims match the runtime handoff.
- For reduced recovery, require data-item provenance, before/after loss, and parameter or optimizer-state changes.
- Confirm that posterior API checks and generated skill invocations were recorded.
- Scan the source manifest for forbidden original-repository paths.
- Confirm that the recovery metric is numeric and tied to the declared target.
Validation
Run:
python /share/project/yuyang/workspace/Paper2Skills/Distiller/skills/module-to-skill/scripts/validate_skill_tree.py /share/project/yuyang/workspace/Paperbench/record/case15/extracted_skills_attempt_001/sbi_toolkit/sbi_recovery_diagnostics --run-tests
For standalone validation of an attempt:
python scripts/recovery_diagnostics.py --attempt-dir /path/to/attempt
Limitations
The diagnostic script checks artifact consistency. It cannot prove that a full sbi package result is scientifically accurate without the underlying executable recovery logs and metrics.
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).