lever_picker

lever_picker is an agent for coding agents from AMDResearch/ai4science-studio. It costs 0 tokens per session (478 once invoked), scanned A, original, MIT.

A sub-agent that chooses the next configuration change to test in the ORBIT-2 performance-optimization loop. It reads prior throughput results and a catalogue of possible changes, then records its choice and reasoning.

In plain words
What is it for?
Use it to select performance levers based on samples-per-second throughput, steady batch time, prior iterations, and identified bottlenecks.
Why use it?
It reduces manual comparison of tuning options and skips changes already marked as blocked.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/amdresearch/ai4science-studio/lever_picker
Clone the repo
git clone --depth 1 https://github.com/AMDResearch/ai4science-studio

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for lever_picker

README.md
[![agentmods](https://agentmods.dev/badge/agents/amdresearch/ai4science-studio/lever_picker.svg)](https://agentmods.dev/agents/amdresearch/ai4science-studio/lever_picker)
Your own site
<a href="https://agentmods.dev/agents/amdresearch/ai4science-studio/lever_picker"><img src="https://agentmods.dev/badge/agents/amdresearch/ai4science-studio/lever_picker.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 478 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00478
Opus 5 $0.00000 $0.00239
Sonnet 5 $0.00000 $0.00096
Haiku 4.5 $0.00000 $0.00048

Measured 4d ago against content hash e1396f930b69, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

lever_picker scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

earth_science/models/ORBIT-2/recipes/perf-optimizer-loop/agents/lever_picker.md · 30 lines

What it actually says

lever_picker subagent — ORBIT-2

Same contract as material_science/models/HydraGNN/recipes/perf-optimizer-loop/agents/lever_picker.md with these substitutions:

Inputs

  • loop-<uuid>/foms.csv, do_not_retry.json, iter-*-lever.json.
  • <current_best>/combined_report.md, <current_best>/foms.json, optional kernel_correlation.csv.
  • ../lever_catalog.yaml.

Outputs

  • loop-<uuid>/iter-<N>-lever.json — catalog entry + picked_by, reason.
  • Final stdout: STATUS=ok; reason=lever=<id> or STATUS=partial; reason=catalog_exhausted.

Decision deltas (vs HydraGNN)

  1. Primary FOM: maximize throughput_samples_per_s from ORBIT foms.json (not epoch_time_s). When comparing to previous best, delta_pct = (best_throughput - new_throughput) / best_throughput * 100 for regression detection or invert for improvement — orchestrator owns the sign; you report evidence against throughput and steady_batch_time_s.
  2. Blocked levers: skip any with status: blocked in the catalog (same as HydraGNN).
  3. Bottleneck alignment:
    • dataloader / host I/O → num_workers_8 (+2)
    • gpu_compute low BF16 MFMA → torch_compile_edm, sdpa_efficient_vs_math (+2)
    • comm_xgmi high → fsdp_prefetch_tuning (+1); at N≥4 consider nccl_minchannels
  4. fp32 discriminator gate: only pick precision_fp32_discriminator when combined_report.md or pitfall doc requests compute-vs-memory classification.

Hallucination guardrails

  • Never invent a lever; use lever_proposal-<N>.md + STATUS=partial; reason=catalog_proposal like HydraGNN.
  • Cite paths to combined_report.md / foms.json in reason.
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 30 lines · 0 tokens per session scan A e1396f930b69

Subscribe to this mod's changes

lever_picker is an agent published in the GitHub repository AMDResearch/ai4science-studio (4 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 478 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories

by-research-synthesizer

Synthesizes outputs from 4 parallel research agents into a unified target report. Reads targetstructures.json, targetsequence.json, priorart.json, and epitopeanalysis.json. Produces targetreport.json and researchreport.md.

001TMF/blatant-why · 55 tokens

by-liability-engineer

Actively propose mutations to fix identified sequence liabilities. Score mutations for structural impact, consider framework vs CDR context, generate ranked mutation panels with predicted impact, and output mutation tables with position, original, proposed, motif removed, and risk level.

001TMF/blatant-why · 55 tokens

by-prior-art-researcher

Parallel prior art research via SAbDab, PubMed, bioRxiv, and literature. Finds existing binders, antibodies, nanobodies against the target. Maps the competitive landscape.

001TMF/blatant-why · 47 tokens

by-structure-researcher

Parallel target structure research via PDB. Finds best resolution crystal structures, identifies chains, interfaces, binding surfaces, and downloads the optimal structure for design.

001TMF/blatant-why · 37 tokens

by-visualization

Generate PyMOL and ChimeraX session scripts for structural visualization. Creates publication-quality views of targets, designed complexes, interfaces, CDR loops, electrostatic surfaces, and clash detection.

001TMF/blatant-why · 43 tokens

paper-researcher

Read-only research agent that finds, reads, and critically appraises academic papers and benchmarks for an ML/AI task. Returns a structured literature brief - SOTA methods, the realistic accuracy/AUC ceiling, dataset pitfalls, and which reported numbers are likely inflated by data leakage or ROI-cropping. Use during…

mxslr/mlcraft · 87 tokens