AGENT SKILL

clip_proxy_recovery

Run a bounded soft-mode CLIP proxy recovery that logs contrastive training and zero-shot evaluation evidence.

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

01 前言 Front Matter

name
clip_proxy_recovery
description
Run a bounded soft-mode CLIP proxy recovery that logs contrastive training and zero-shot evaluation evidence.

02 资源文件 2

03 SKILL.md 正文

CLIP Proxy Recovery

Use this skill when full web-scale CLIP pretraining is blocked but soft-mode recovery permits a declared proxy. Do not use it as proof of full CLIP performance.

Inputs

  • Distiller attempt directory with module_plan.json and generated skills.
  • Runtime handoff documenting package/model blockers.
  • Tiny paired feature fixture or generated proxy data item.

Outputs

  • recovery_result.json with numeric proxy metrics.
  • generated_data_item.json, training_trace.json, experiment command log, and generated-skill invocation log.
  • Mechanism checks for pair construction, contrastive loss, optimizer update, prompt classification, and source boundary.

Workflow

  1. Read the module-plan target and runtime handoff.
  2. Create or validate a tiny paired image-text feature dataset.
  3. Run a real optimizer update on trainable scalar parameters using symmetric contrastive loss.
  4. Classify held-out image vectors with prompt-derived class vectors.
  5. Write auditable logs and mark the recovery as proxy/reduced.

Validation

Run the recovery harness and then validate_recovery_experiment.py on the attempt directory.

Limitations

This skill validates the mechanism, not the paper's full ImageNet-scale accuracy.

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