Exécutez n'importe quel Skill dans Manus
en un clic

Exécutez n'importe quel Skill dans Manus en un clic

helios-real-workload-ab

Étoiles2

Forks0

Mis à jour30 juin 2026 à 22:59

Build and run Helios real-context A/B benchmarks with deterministic workload manifests, correctness gates, and decision artifacts.

Installation

Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.

Exécuter dans Manus

Source

jscott3201

jscott3201/Helios

Ouvrir le dépôt GitHub Voir les dépôts du créateur

Téléchargement

Exécuter dans Manus

Métiers associésSOC

Basé sur la classification professionnelle SOC

Développeurs de logicielsProfessions informatiques et mathématiques·SOC 15-1252

SKILL.md

readonly

name	helios-real-workload-ab
description	Build and run Helios real-context A/B benchmarks with deterministic workload manifests, correctness gates, and decision artifacts.

Helios real-workload A/B skill

Use this skill when a task mentions Helios performance, A/B benchmarking, real workloads, benchmark corpus, repeated-prefix testing, or tiny16 performance decisions.

Procedure

Read BENCHMARKS.md, docs/xr-real-workload-methodology.md, and docs/xr-ab-methodology.md.
Identify the current baseline and candidate before editing code.
Prefer repo-local real contexts over repeated-token prompts.
Record exact commands, git SHA, model identity, env vars, and output paths.
Use at least records.jsonl, summary.json, report.md, blockers.md, and decision.md for every Goal.
Do not accept performance gains without correctness and memory gates.
Mark low-trial or high-variance results honestly.

Decision labels

Use exactly one:

accept_candidate
reject_candidate
keep_experimental
needs_more_data
blocked_with_evidence

Plus depuis ce dépôt

même dépôt

helios-mtp-acceptance-analysis

jscott3201/Helios

Diagnose Gemma 4 MTP acceptance and speculative decoding behavior in Helios using real-context prompts, trace records, and exactness gates.

2026-06-302

helios-native-mlx-ab-profiling

jscott3201/Helios

Profile and optimize Helios native MLX/C++ execution paths with A/B evidence, memory gates, and feature-flagged experimental kernels.

2026-06-302

gemma4d-16gb-mac-profiling

jscott3201/Helios

Use for tiny16 MacBook profiling: memory pressure, context-length limits, TTFT/decode metrics, cache effectiveness, MTP acceptance, and release-gate benchmark reports.

2026-06-302

gemma4d-dynamic-adapters

jscott3201/Helios

Use for dynamic LoRA/QLoRA adapter import, validation, loading, routing, adapter-aware KV cache keys, aLoRA planning, and adapter memory residency.

2026-06-302

gemma4d-kv-cache-offload

jscott3201/Helios

Use for Gemma4D KV cache, prefix cache, SSD cold tier, cache manifests, cache-key correctness, quantized/compressed KV experiments, and restore parity.

2026-06-302

gemma4d-milestone-execution

jscott3201/Helios

Use to execute one Gemma4D milestone end-to-end with scoped tasks, verification commands, evidence capture, and blocker reporting. Trigger on milestone implementation, Codex Goals, task planning, or release gates.

2026-06-302

name	helios-real-workload-ab
description	Build and run Helios real-context A/B benchmarks with deterministic workload manifests, correctness gates, and decision artifacts.

Helios real-workload A/B skill

Use this skill when a task mentions Helios performance, A/B benchmarking, real workloads, benchmark corpus, repeated-prefix testing, or tiny16 performance decisions.

Procedure

Read BENCHMARKS.md, docs/xr-real-workload-methodology.md, and docs/xr-ab-methodology.md.
Identify the current baseline and candidate before editing code.
Prefer repo-local real contexts over repeated-token prompts.
Record exact commands, git SHA, model identity, env vars, and output paths.
Use at least records.jsonl, summary.json, report.md, blockers.md, and decision.md for every Goal.
Do not accept performance gains without correctness and memory gates.
Mark low-trial or high-variance results honestly.

Decision labels

Use exactly one:

accept_candidate
reject_candidate
keep_experimental
needs_more_data
blocked_with_evidence