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.
npx skills add khalilbenaz/claude-skills-collection --skill model-optimization-guidegit clone --depth 1 https://github.com/khalilbenaz/claude-skills-collectionWrote 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.
[](https://agentmods.dev/skills/khalilbenaz/claude-skills-collection/model-optimization-guide)<a href="https://agentmods.dev/skills/khalilbenaz/claude-skills-collection/model-optimization-guide"><img src="https://agentmods.dev/badge/skills/khalilbenaz/claude-skills-collection/model-optimization-guide/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/khalilbenaz/claude-skills-collection/model-optimization-guide"><img src="https://agentmods.dev/badge/skills/khalilbenaz/claude-skills-collection/model-optimization-guide.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00126 | $0.02716 |
| Opus 5 | $0.00063 | $0.01358 |
| Sonnet 5 | $0.00025 | $0.00543 |
| Haiku 4.5 | $0.00013 | $0.00272 |
Grade A, and why
model-optimization-guide 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 9d 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.
How it starts
The opening of the file, as written. The whole thing — 300 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Model Optimization Guide
Guide opérationnel pour réduire la taille et accélérer l'inférence d'un modèle ML sans régresser les métriques métier.
Critères de décision rapides
| Contrainte principale | Technique recommandée |
|---|---|
| Déploiement sans GPU, CPU seul | PTQ INT8 + ONNX Runtime (OpenVINO EP) |
| Latence < 20 ms sur GPU NVIDIA | TensorRT FP16 ou INT8 |
| Modèle trop gros pour mémoire edge | Pruning structuré + TFLite/CoreML |
| LLM > 7B à servir sur 1 GPU | GPTQ 4-bit ou AWQ |
| Contrainte de qualité stricte (< 0,5 % drop) | QAT ou distillation |
| Pipeline cross-framework | ONNX export + ORT |
Workflow en 7 étapes
1. Profiler et fixer les objectifs
Avant toute optimisation, mesurer la baseline sur le hardware de production.
# PyTorch Profiler (CPU+GPU)
import torch
from torch.profiler import profile, ProfilerActivity
with profile(activities=[ProfilerActivity.CPU, ProfilerActivity.CUDA],
record_shapes=True, profile_memory=True) as prof:
model(inputs)
print(prof.key_averages().table(sort_by="cuda_time_total", row_limit=15))
prof.export_chrome_trace("trace.json") # ouvrir dans chrome://tracing
Grille d'objectifs à remplir avant de commencer :
| Métrique | Valeur actuelle | Cible | Tolérance dégradation |
|---|---|---|---|
| Latence P95 (ms) | ? | ? | +0 % |
| Taille modèle (MB) | ? | ? | — |
| Accuracy / F1 | ? | — | -0,5 % max |
| VRAM / RAM (MB) | ? | ? | — |
2. Quantification (technique la plus rapide)
PTQ INT8 — PyTorch (sans réentraînement)
import torch.quantization as tq
model.eval()
model.qconfig = tq.get_default_qconfig('fbgemm') # CPU x86
# ou 'qnnpack' pour ARM/mobile
tq.prepare(model, inplace=True)
# Calibration : passer ~100-500 exemples représentatifs
for batch in calibration_loader:
model(batch)
tq.convert(model, inplace=True)
torch.save(model.state_dict(), "model_int8.pt")
PTQ FP16 — ONNX Runtime (le plus portable)
from onnxruntime.quantization import quantize_dynamic, QuantType
quantize_dynamic(
"model.onnx",
"model_int8.onnx",
weight_type=QuantType.QInt8
)
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.
- 9d ago First seen · 300 lines · 126 tokens per session scan A 1c7af8cfa789
model-optimization-guide is a skill published in the GitHub repository khalilbenaz/claude-skills-collection (22 stars, last pushed 15d ago), licensed MIT. It adds 126 tokens to every session and 2,716 once invoked, about $0.0006 per session on Opus 5. 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 2026-08-30.
Other skills, from other repositories
edgartools
Python library for accessing, analyzing, and extracting data from SEC EDGAR filings. Use when working with SEC filings, financial statements (income statement, balance sheet, cash flow), XBRL financial data, insider trading (Form 4), institutional holdings (13F), company financials, annual/quarterly reports (10-K…
rag-patterns
RAG: embeddings, chunking, hybrid search (BM25+vector), reranking, CRAG, multi-hop. Triggers: RAG, embedding, pgvector, Qdrant, Pinecone, Weaviate, reranker, semantic search.
model-routing-patterns
Multi-model pipelines (Haiku/Sonnet/Opus): cost routing, escalation, fallback chains. Triggers: model routing, Haiku, Sonnet, Opus, escalation, fallback chain.
evaluate
Evaluates RAG retrieval and LLM-as-judge metrics (faithfulness, relevancy, context precision). Triggers: measure RAG quality, knowledge gap, RAG eval, golden dataset.
json-mode-patterns
Structured JSON output from Claude: tool-use-as-JSON, schema, parsing, partial recovery. Triggers: JSON mode, structured output, schema validation, JSON parsing.
explain
Explains code/architecture with Mermaid diagrams and sequence flows. Triggers: what does X do, how does Y work, explain code, sequence diagram.