AGENT SKILL

ptq4vit_recovery_harness

Use this skill to run a bounded PTQ4ViT proxy recovery that invokes generated scale-search, twin-quantization, and Hessian-metric skills.

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

01 前言 Front Matter

name
ptq4vit_recovery_harness
description
Use this skill to run a bounded PTQ4ViT proxy recovery that invokes generated scale-search, twin-quantization, and Hessian-metric skills.

02 资源文件 2

03 SKILL.md 正文

Ptq4Vit Recovery Harness

Use this skill when a recovery or implementation task needs the PTQ4ViT capability described in the module plan. Do not use it to claim full ImageNet accuracy without an actual model, dataset, and evaluation command.

Inputs

  • Numeric calibration tensors or an attempt directory, depending on the script.
  • Explicit bit width, candidate scales, and runtime/source-boundary constraints.

Outputs

  • Deterministic numeric traces suitable for validation.
  • JSON-compatible metadata that downstream recovery can inspect.

Workflow

  1. Read the module contract and pass only bounded local tensors or attempt artifacts.
  2. Run the script in scripts/ or import its pure functions.
  3. Validate outputs with the tests in tests/.
  4. Record commands and artifacts in the Distiller attempt when used for recovery.

Validation

Run python validate_skill_tree.py <skill_dir> --run-tests from the Distiller module-to-skill validator.

Limitations

This skill preserves the PTQ4ViT mechanism but does not bundle pretrained models, ImageNet, or the original repository. Recovery must not read the original repository.

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