proxy_training_recovery
Run a bounded Moderate-DS proxy recovery harness with generated skill invocations and optimizer evidence.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- proxy_training_recovery
- description
- Run a bounded Moderate-DS proxy recovery harness with generated skill invocations and optimizer evidence.
02 资源文件 2
- 脚本scripts/run_proxy_recovery.py6.2 KB
- 脚本tests/test_proxy.py427 B
03 SKILL.md 正文
Proxy Training Recovery Harness
Use this skill when full image-dataset retraining is blocked but soft-mode recovery permits a declared mechanism-faithful proxy. The harness must import or call the scoring, median-selection, and ablation skills, produce executable logs, and write validator-compatible recovery artifacts.
Inputs
- Attempt directory containing
module_plan.jsonandenvironment/runtime_handoff.json. - Generated skills root containing the Moderate-DS module skills.
- Optional synthetic dataset parameters for stress tests.
Outputs
recovery/recovery_result.jsonwith numeric metrics and mechanism checks.recovery/logs/generated_data_item.jsondescribing the proxy data.recovery/logs/training_trace.jsonshowing loss before and after a real optimizer update.recovery/logs/generated_skill_invocations.jsonproving all core skills were exercised.
Workflow
- Create a deterministic class-cluster representation dataset with central, moderate, and far points.
- Import the generated scoring, median-selection, and ablation scripts from the generated skills root.
- Select the moderate coreset and compare it against extreme policies.
- Run a simple trainable linear proxy update on the selected coreset and log parameter changes.
- Write recovery artifacts and source boundaries before invoking validation.
Validation
Run python scripts/run_proxy_recovery.py --attempt-dir <attempt_dir> --skills-root <skills_root>. Then run the Distiller recovery validator on the attempt directory.
Limitations
This is not a full CIFAR-100/ImageNet reproduction. It is a soft-mode proxy that validates the mechanism under bounded runtime constraints.
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).