AGENT SKILL

adaptive_apgd_loop

Run a deterministic APGD-style projected attack with loss-history step adaptation and auditable traces.

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

01 前言 Front Matter

name
adaptive_apgd_loop
description
Run a deterministic APGD-style projected attack with loss-history step adaptation and auditable traces.

02 资源文件 2

03 SKILL.md 正文

Adaptive APGD Loop

Use this skill to recover or test the APGD mechanism from AutoAttack when a bounded, deterministic implementation is needed. Do not treat this pure-Python finite-difference implementation as a high-throughput replacement for framework autograd attacks.

Inputs

  • Logit function accepting one vector.
  • Examples and labels.
  • Projection and loss helpers.
  • Norm, epsilon, iterations, initial step size, and adaptation window.

Outputs

  • Adversarial examples and predictions.
  • Per-example success flags.
  • Loss trace and step-size adaptation events.

Workflow

  1. Start from the clean example or a deterministic offset inside the threat ball.
  2. Estimate the loss gradient with finite differences.
  3. Take an ascent step and project back to the threat model.
  4. Track the best loss and adversarial point.
  5. On non-improving windows, halve step size and continue from the best point.

Validation

Run python tests/test_apgd.py or the skill-tree validator.

Limitations

Finite differences are only for compact recovery evidence. Production use should substitute exact gradients while preserving the same adaptation and projection contracts.

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