AGENT SKILL

lora_recovery_evaluation

Assemble and validate a soft-mode reduced LoRA recovery with executable mechanism evidence and source-boundary logs.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
lora_recovery_evaluation
description
Assemble and validate a soft-mode reduced LoRA recovery with executable mechanism evidence and source-boundary logs.

02 资源文件 2

03 SKILL.md 正文

LoRA Recovery Evaluation Harness

Use this skill after LoRA module skills and runtime probes are available. It packages a reduced or full recovery into auditable artifacts: experiment plan, command log, generated skill invocation log, source manifest, recovery result, and validator output. Do not use the original implementation repository during recovery.

Inputs

  • Attempt directory with module_plan.json and environment/runtime_handoff.json.
  • Generated skills root containing the LoRA layer, freezing, and training-step skills.
  • Recovery mode and the declared full or proxy target.

Outputs

  • Executable recovery command evidence.
  • Numeric metric aligned with module_plan.fast_recovery_target.
  • Mechanism checks for frozen weights, low-rank update, optimizer step, checkpoint filtering, and merge equivalence.

Workflow

  1. Read the module plan target and runtime handoff first.
  2. If full torch/transformers recovery is unavailable, declare the soft-mode reduced target before running the harness.
  3. Invoke generated skill scripts rather than duplicating their contracts silently.
  4. Save training trace, generated data item, command log, source manifest, and recovery result.
  5. Run the official recovery validator and treat failures as refinement input.

Validation

Run python scripts/evaluate_recovery.py --attempt-dir <attempt> --skills-root <skills>. The command calls the reduced training skill and reports a numeric loss-reduction ratio.

Limitations

This harness validates mechanism fidelity for reduced recovery. It cannot certify paper-table metrics without real packages, pretrained models, and benchmark datasets.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

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).

★ 351