Skip to content
Skills Library
AI & agents · Conversion & formats

llama-cpp-gguf-quantization

/llama-cpp-gguf-quantization

Use this skill when converting, quantizing, and running models on consumer CPUs and Apple Silicon hardware using llama.cpp.

Licence
MIT
Last repository change
28.10.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 28.10.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Use this skill when converting, quantizing, and running models on consumer CPUs and Apple Silicon hardware using llama.cpp.

Original author description

Use this skill when converting, quantizing, and running models on consumer CPUs and Apple Silicon hardware using llama.cpp.

How it works

  1. When to use
  2. Prerequisites
  3. Step-by-step workflow
  4. Commands / code examples
  5. Common mistakes & pitfalls
  6. Verification checklist
  7. Related skills
Original SKILL.md headings · 45 lines, 264 words
SKILL.md preview
---
name: llama-cpp-gguf-quantization
description: Use this skill when converting, quantizing, and running models on consumer CPUs and Apple Silicon hardware using llama.cpp.
---
# GGUF Model Quantization & llama.cpp

## When to use
Use this skill when you need to converting, quantizing, and running models on consumer CPUs and Apple Silicon hardware using llama.cpp.

## Prerequisites
- Deep understanding of modern AI/ML systems, data engineering pipelines, or distributed architectures.
- Testing environment with necessary libraries and credentials configured.

## Step-by-step workflow
1. **Analyze data & architectural requirements**: Evaluate throughput, latency, context windows, and schema constraints.
2. **Formulate pipeline / model contracts**: Define strict input/output specifications and validation schemas.
3. **Implement core execution logic**: Construct deterministic transformations, prompt chains, or streaming data flows.
4. **Validate performance & accuracy**: Run automated test assertions, evaluate accuracy metrics, and benchmark throughput.
5. **Enforce observability & guardrails**: Instrument workflows with tracing, logging, rate limits, and fallback strategies.

## Commands / code examples
```python
# Production pipeline and implementation pattern for llama-cpp-gguf-quantization
import os

def run_pipeline():
    print("Executing operational pipeline: llama-cpp-gguf-quantization")

if __name__ == "__main__":
    run_pipeline()
```

## Common mistakes & pitfalls
- Allowing untrusted inputs to bypass validation checks, resulting in data corruption or prompt injection.
- Accumulating unconstrained state in memory during stream processing, leading to Out-Of-Memory (OOM) crashes.
- Neglecting evaluation benchmarks, deploying models or prompts without quantitative quality scoring.

## Verification checklist
- [ ] Pipeline validates input schemas and filters malformed records.
- [ ] Concurrency and memory utilization remain within provisioned boundaries.
- [ ] Performance and accuracy metrics meet predefined SLA targets.
- [ ] Fallback and retry mechanisms handle external API or network interruptions gracefully.

## Related skills
- retrieval-augmented-generation-rag-pipeline
- prompt-injection-defense-guardrails
- apache-kafka-event-streaming-architecture
- distributed-tracing-opentelemetry-sdk

Versions

The database has no other skills named llama-cpp-gguf-quantization.

Database snapshot: 13.11.2026 · package version 574897612f39 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100minicpm5-deploy-llama-cpp
    Run MiniCPM5-1B or MiniCPM5-2B with llama.cpp using the released GGUF artifacts (F16 / Q8_0 / Q4_K_M). Use when the user wants CPU-only / consumer-GPU / cross-platform native deployment, asks for "llama.cpp", "llama-cli", "llama-server", "GGUF", or has no Python available.
  • 84llama-cpp
    Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.
  • 95llama-cpp
    Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.
  • 100gemma4-local-deploy
    在本机 Mac 或 Apple Silicon 上部署 Gemma 4 12B。本地安装/升级 llama.cpp,下载 GGUF 量化模型,用 llama-server 暴露 OpenAI-compatible API,或用 Ollama 暴露本地模型服务;按用户需求在默认 Q4_K_M、64K/128K 长上下文、QAT Q4_0 @ 256K、左右对比演示之间选择,配置 tmux 后台运行,验证健康检查、问答接口、资源占用和常见故障。当用户说部署 Gemma 4、Gemma 4…
  • 100llama-cpp
    Guide for llama.cpp, the C/C++ LLM inference framework by ggml-org. Covers the C API (llama.h), GGUF format, quantization (Q4_K_M, Q8_0, IQ4_XS), CMake builds, GPU backends (CUDA, Vulkan, Metal, ROCm), HTTP server with OpenAI-compatible API, embeddings, grammar constraints,…
  • 100llama-cpp
    Run GGUF models directly, load LoRA adapters, benchmark inference speed, and serve models via llama-server using llama.cpp. Includes Qwen 3.5 serve scripts (9B dense + F16, 35B MoE) with asymmetric KV cache and thinking mode. Secondary to Ollama; use when needing direct model…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.