Text Generation
GGUF
mazz-code
aengenium
code
reasoning
weight-space-programming
model-surgery
vision
video
abliterated
conversational
Instructions to use mazz92/mazz-models with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use mazz92/mazz-models with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf mazz92/mazz-models:Q4_K_M # Run inference directly in the terminal: llama cli -hf mazz92/mazz-models:Q4_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf mazz92/mazz-models:Q4_K_M # Run inference directly in the terminal: llama cli -hf mazz92/mazz-models:Q4_K_M
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf mazz92/mazz-models:Q4_K_M # Run inference directly in the terminal: ./build/bin/llama cli -hf mazz92/mazz-models:Q4_K_M
Use Docker
docker model run hf.co/mazz92/mazz-models:Q4_K_M
- Llama new
- LM Studio
- Jan
- vLLM
How to use mazz92/mazz-models with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "mazz92/mazz-models" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "mazz92/mazz-models", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/mazz92/mazz-models:Q4_K_M
- Ollama
How to use mazz92/mazz-models with Ollama:
ollama run hf.co/mazz92/mazz-models:Q4_K_M
- Unsloth Desktop
- Pi
How to use mazz92/mazz-models with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf mazz92/mazz-models:Q4_K_M
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:9931/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "mazz92/mazz-models:Q4_K_M" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use mazz92/mazz-models with Docker Model Runner:
docker model run hf.co/mazz92/mazz-models:Q4_K_M
- Lemonade
How to use mazz92/mazz-models with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull mazz92/mazz-models:Q4_K_M
Run and chat with the model
lemonade run user.mazz-models-Q4_K_M
List all available models
lemonade list
- Hermes Agent
How to use mazz92/mazz-models with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf mazz92/mazz-models:Q4_K_M
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:9931/v1 hermes config set model.default mazz92/mazz-models:Q4_K_M
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use mazz92/mazz-models with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf mazz92/mazz-models:Q4_K_M
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:9931/v1 \ --custom-model-id "mazz92/mazz-models:Q4_K_M" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
|
Download README.md from mazz92/mazz-models: direct link, hf CLI and curl.
- Browser
- Download file 7.42 kB
-
https://huggingface.co/mazz92/mazz-models/resolve/main/README.md
- Command line
-
hf download hf://mazz92/mazz-models/README.md
-
curl -L -o README.md https://huggingface.co/mazz92/mazz-models/resolve/main/README.md
7.42 kB
| license: apache-2.0 | |
| tags: | |
| - mazz-code | |
| - aengenium | |
| - code | |
| - reasoning | |
| - weight-space-programming | |
| - model-surgery | |
| - vision | |
| - video | |
| - abliterated | |
| - gguf | |
| pipeline_tag: text-generation | |
| # MAZZ รNGENIUM โ Flotte de Modรจles Souverains & Chirurgicaux | |
| Bienvenue sur le dรฉpรดt officiel de la suite neuronale souveraine **MAZZ รNGENIUM**. | |
| * **Auteur & Concepteur** : Eren Tipirdamaz | |
| * **Organisation** : MAZZ รNGENIUM (*Living Engineering : le vivant inspire, l'ingรฉnierie donne forme*) | |
| * **Application hรดte** : MAZZ Code ([https://github.com/mazzadev92/mazz-code](https://github.com/mazzadev92/mazz-code)) | |
| * **Licence** : Apache-2.0 | |
| --- | |
| ## ๐๏ธ Cartographie de la Flotte Souveraine & รditions Chirurgicales | |
| ``` | |
| โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ | |
| โ MAZZ-TECHNOS-APEXCODER-70B-SURGICAL (39.60 Go) โ | |
| โ Fusion Arcee Nemotron-70B + Llama-3.3-70B Abliterated โ | |
| โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ | |
| โ | |
| โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ | |
| โ โ | |
| โโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ | |
| โ MAZZ-TECHNOS-TITANCODER-32B-SURGICAL (18.5Go) โ โ MAZZ-TECHNOS-CODER-32B & TITAN-32B-COT (18.49 Go) โ | |
| โ Fusion 90/10 DeepSeek-R1 + Qwen2.5-Coder โ โ Champions Infรฉrence & Chaรฎne de Pensรฉe Unitaire โ | |
| โโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ | |
| โ โ | |
| โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโค | |
| โ โ โ | |
| โโโโโโโโโโโโโโผโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโผโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโผโโโโโโโโโโโโโโโ | |
| โ MAZZ-TECHNOS-BASE-4B โ โ MAZZ-TECHNOS-VISION-FLUX โ โ MAZZ-TECHNOS-MOTION-WAN โ | |
| โ Polyvalent CPU (2.52 Go) โ โ Gรฉnรฉration Image (6.31 Go)โ โ Gรฉnรฉration Vidรฉo (1.43 Go)โ | |
| โโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโ | |
| โ | |
| โโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ | |
| โ MAZZ-TECHNOS-DRAFT-0.8B-SURGICAL (0.76 Go) โ | |
| โ ๐ฌ Weight-Space Programming MIT CSAIL (Down-Proj Patching)โ | |
| โ MATH: 7.9% โ 48.1% | GSM8K: 13.8% โ 67.1% | Zรฉro Boucle โ | |
| โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ | |
| ``` | |
| --- | |
| ### ๐ฌ La Gamme Chirurgicale MAZZ (Model Surgery & Weight-Space Programming) | |
| 1. **`Mazz-Technos-Draft-0.8B-Surgical.Q8_0.gguf`** (0,76 Go) | |
| * **Base** : Qwen3.5-0.8B Abliterated. | |
| * **Innovation** : Intรจgre la mรฉthode rรฉvolutionnaire du MIT CSAIL (*Sparks of Weight-Space Programming*, Takehi & Khattab). | |
| * **Intervention** : Modification ciblรฉe de 64 colonnes dans les matrices `down_proj` des couches 15 et 19 (0,016% des paramรจtres รฉditรฉs sans gradient descent). | |
| * **Gains** : MATH-500 (7,9 % โ 48,1 %), GSM8K (13,8 % โ 67,1 %), IFEval (21,0 % โ 62,3 %), blocages en boucle rรฉduits de 93,0 % ร 33,1 %. | |
| 2. **`Mazz-Technos-TitanCoder-32B-Surgical.Q4_K_M.gguf`** (18,49 Go) | |
| * **Base** : Fusion chirurgicale pondรฉrรฉe 90/10 entre **DeepSeek-R1-Distill-Qwen-32B** (capacitรฉ de raisonnement pur et rรฉflexion interne `<thought>`) et **Qwen2.5-Coder-32B-Instruct** (synthรจse de code, refactoring et complรฉtion AST). | |
| * **Caractรฉristiques** : Raisonnement dรฉductif autonome combinรฉ ร une prรฉcision d'exรฉcution de code chirurgicale, entiรจrement non censurรฉ. | |
| 3. **`Mazz-Technos-ApexCoder-70B-Surgical.Q4_K_M.gguf`** (39,60 Go) | |
| * **Base** : Fusion chirurgicale Arcee entre **Nvidia Nemotron-70B** et **Meta Llama-3.3-70B Abliterated**. | |
| * **Caractรฉristiques** : Le sommet absolu de la flotte MAZZ. Analyse contextuelle massive, architecture logicielle distribuรฉe, audits de sรฉcuritรฉ et refactorings de dรฉpรดts complexes. | |
| --- | |
| ### ๐ข Tier 1 : Grand Public & CPU Local (Edge Sovereign) | |
| * **`Mazz-Technos-Draft-0.8B-Surgical.Q8_0.gguf`** (0,76 Go) : Modรจle chirurgical ultra-rapide (>40 tok/s), raisonnement aiguisรฉ sans boucles. | |
| * **`Mazz-Technos-Base-4B.Q4_K_M.gguf`** (2,52 Go) : Polyvalent grand public pour PC portables et CPU standards (4-6 tok/s). | |
| --- | |
| ### ๐ต Tier 2 : Fleurons Ingรฉnierie & Raisonnement (Station de Travail) | |
| * **`Mazz-Technos-TitanCoder-32B-Surgical.Q4_K_M.gguf`** (18,49 Go) : Fusion chirurgicale Raisonnement R1 + Ingรฉnierie Coder. | |
| * **`Mazz-Technos-Coder-32B.Q4_K_M.gguf`** (18,49 Go) : Champion du code, refactoring chirurgical, tool-calling et zรฉro refus. | |
| * **`Mazz-Technos-Titan-32B-CoT.Q4_K_M.gguf`** (18,49 Go) : Raisonnement scientifique profond avec blocs `<thought>` (Chain-of-Thought). | |
| --- | |
| ### ๐ฃ Tier 3 : Crรฉation Multimodale Souveraine | |
| * **`Mazz-Technos-Vision-Flux.Q4_0.gguf`** (6,31 Go) : Gรฉnรฉration d'images photorรฉalistes en 4 รฉtapes (FLUX.1-schnell 12B). | |
| * **`Mazz-Technos-Motion-Wan.Q8_0.gguf`** (1,43 Go) : Gรฉnรฉration vidรฉo cinรฉmatique fluide (Wan 2.1 1.3B). | |
| --- | |
| ### ๐ด Tier 4 : Sommet Cognitif & Architecture | |
| * **`Mazz-Technos-ApexCoder-70B-Surgical.Q4_K_M.gguf`** (39,60 Go) : Le supercalculateur cognitif pour l'analyse de dรฉpรดts volumineux. | |
| --- | |
| ## ๐ป Intรฉgration dans MAZZ Code | |
| ```bash | |
| # Lancement direct via MAZZ CLI : | |
| mazz start --model Mazz-Technos-TitanCoder-32B-Surgical.Q4_K_M.gguf | |
| ``` | |
| --- | |
| *Conรงu avec rigueur par Eren Tipirdamaz pour l'รฉcosystรจme MAZZ รNGENIUM.* | |