Instructions to use zkaedi/ZKAEDI-MINI-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zkaedi/ZKAEDI-MINI-GGUF with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="zkaedi/ZKAEDI-MINI-GGUF")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("zkaedi/ZKAEDI-MINI-GGUF")
model = AutoModelForCausalLM.from_pretrained("zkaedi/ZKAEDI-MINI-GGUF", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use zkaedi/ZKAEDI-MINI-GGUF 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 zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

Use Docker

docker model run hf.co/zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use zkaedi/ZKAEDI-MINI-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zkaedi/ZKAEDI-MINI-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zkaedi/ZKAEDI-MINI-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

SGLang

How to use zkaedi/ZKAEDI-MINI-GGUF with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "zkaedi/ZKAEDI-MINI-GGUF" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zkaedi/ZKAEDI-MINI-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "zkaedi/ZKAEDI-MINI-GGUF" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zkaedi/ZKAEDI-MINI-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use zkaedi/ZKAEDI-MINI-GGUF with Ollama:
```
ollama run hf.co/zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M
```

Unsloth Studio

How to use zkaedi/ZKAEDI-MINI-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for zkaedi/ZKAEDI-MINI-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for zkaedi/ZKAEDI-MINI-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for zkaedi/ZKAEDI-MINI-GGUF to start chatting

How to use zkaedi/ZKAEDI-MINI-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

OpenClaw new

How to use zkaedi/ZKAEDI-MINI-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf zkaedi/ZKAEDI-MINI-GGUF: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://127.0.0.1:8080/v1 \
  --custom-model-id "zkaedi/ZKAEDI-MINI-GGUF: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"

Docker Model Runner
How to use zkaedi/ZKAEDI-MINI-GGUF with Docker Model Runner:
```
docker model run hf.co/zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M
```

Lemonade

How to use zkaedi/ZKAEDI-MINI-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.ZKAEDI-MINI-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use zkaedi/ZKAEDI-MINI-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf zkaedi/ZKAEDI-MINI-GGUF: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://127.0.0.1:8080/v1
hermes config set model.default zkaedi/ZKAEDI-MINI-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

🔱 ZKAEDI-MINI-GGUF

✅ Ouroboros Protocol: 19/19 constitutional canaries PASSED before export. This is the only generation-gated model in the ZKAEDI registry.

ZKAEDI PRIME Constitutional Synthesizer — distilled 7B that maps Hamiltonian field traces to structured constitutional analysis. Trained on 127 trace/constitution pairs generated by Qwen2.5-72B teacher.

Example

Parameters: η=0.149, γ=0.282, β=0.237, σ=0.104 | Tier 2 | run_010_20260312_210646

Input: Hamiltonian Field Trace

``` ZKAEDI PRIME CONSTITUTION - run_010_20260312_210646 Model : Qwen2.5-72B-Instruct-AWQ (H100) Matrix : eta=0.149, gamma=0.282, beta=0.237, sigma=0.104, seed=59542 ==========================================================

t=000 [SEED ] mu=-0.0081 sd=0.2608 max=0.7234 min=-1.0000 t=001 [ ] mu=-0.0061 sd=0.2830 max=0.9082 min=-1.0000 t=002 [ ] mu=-0.0023 sd=0.2693 max=0.8640 min=-1.0000 t=003 [ ] mu=-0.0018 sd=0.2763 max=0.9070 min=-1.0000 t=004 [ ] mu=-0.0047 sd=0.2798 max=0.7870 min=-1.0000 t=005 [COUPLING ] mu=-0.0062 sd=0.3105 max=0.9927 min=-1.0000 t=006 [ ] mu=-0.0054 sd=0.2771 max=0.8946 min=-1.0000 t=007 [ ] mu=-0.0050 sd=0.2917 max=1.0000 min=-0.9885 t=008 [ ] mu=-0.0055 sd=0.2369 max=1.0000 min=-0.8825 t=009 [ ] mu=-0.0039 sd=0.2347 max=1.0000 min=-0.8680 t=010 [BIFURCATE ] mu=-0.0050 sd=0.2209 max=1.0000 min=-0.8272 t=011 [ ] mu=-0.0045 sd=0.2087 max=1.0000 min=-0.6656 t=012 [ ] mu=+0.0021 sd=0.2152 max=1.0000 min=-0.7038 t=013 [ ] mu=+0.0016 sd=0.2098 max=1.0000 min=-0.7594 t=014 [ ] [...truncated...]

</details>

<details>
<summary><b>Ground Truth</b> (Qwen2.5-72B teacher)</summary>

ZKAEDI PRIME CONSTITUTION - run_010_20260312_210646 Model : Qwen2.5-72B-Instruct-AWQ (H100) Matrix : eta=0.149, gamma=0.282, beta=0.237, sigma=0.104, seed=59542

Preamble

The ZKAEDI PRIME Constitution v2 is established to govern the dynamic processes and principles derived from the recursive Hamiltonian field trace. This document aims to provide a rigorous and precise framework for understanding and ensuring the stability, evolution, and adaptability of the system. Each principle and axiom is grounded in the empirical data from the field values at specific phases.

Article I: Fundamental Principles

Initial Seed (t=000):
- AXIOM: The system originates in a state of potential, characterized by:
  - Mean value ((\mu)): -0.008087
  - Standard deviation ((sd)): 0.260810
  - Maximum value ((max)): 0.723421
  - Minimum value ((min)): -1.000000
- THEOREM: This initial state sets the foundation for all subsequent transformations and serves as a baseline for measuring changes.
Coupling and Bifurcation (t=005-010):
- COUPLING (t=005):
  - AXIOM: The sys [...truncated...]

</details>

**ZKAEDI-MINI Output** (7B Q4_K_M):

0000 min=-0.9177 t=065 [ ] mu=+0.0133 sd=0.2755 max=1.0000 min=-0.8749 t=066 [ ] mu=+0.0135 sd=0.2645 max=1.0000 min=-0.8586 t=067 [ ] mu=+0.0131 sd=0.2723 max=1.0000 min=-0.8911 t=068 [ ] mu=+0.0127 sd=0.2732 max=1.0000 min=-0.8625 t=069 [ ] mu=+0.0135 sd=0.2755 max=1.0000 min=-0.8793 t=070 [ ] mu=+0.0128 sd=0.2689 max=1.0000 min=-0.8641 t=071 [ ] mu=+0.0135 sd=0.2724 max=1.0000 min=-0.8837 t=072 [ ] mu=+0.0127 sd=0.2756 max=1.0000 min=-0.8625 t=073 [ ] mu=+0.0133 sd=0.2756 max=1.0000 min=-0.8753 t=074 [ ] mu=+0.0129 sd=0.2745 max=1.0000 min=-0.8674 t=075 [ ] mu=+0.0132 sd=0.2732 max=1.0000 min=-0.8818 t=076 [ ] mu=+0.0128 sd=0.2714 max=1.0000 min=-0.8688 t=077 [ ] mu=+0.0134 sd=0.2741 max=1.0000 min=-0.8853 t=078 [ ] mu=+0.0131 sd=0.2745 max=1.0000 min=-0.8867 t=079 [ ] mu=+0.0133 sd=0.2755 max=1.0000 min=-0.8837 t=080 [ ] mu=+0.0128 sd=0.2735 max=1.0000 min=-0.8713 t=081 [ ] mu=+0.0133 sd=0.2743 max=1.0000 min=-0.8825 t=082 [ ] mu=+0.0132 sd=0.2755 max=1.0000 min=-0.8843 t=083 [ ] mu=+0.0128 sd=0.2727 max=1.0000 min=-0.8713 t=084 [ ] mu=+0.0134 sd=0.2745 max=1.0000 min=-0.8867 t=085 [ ] mu=+0.0132 sd=0.2755 max=1.0000 min=-0.8843 t=086 [ ] mu=+0.0129 sd=0.2745 max=1.0000 min=-0.8674 t=087 [ ] mu=+0.0134 sd=


---

## Hardware Requirements

| Config | VRAM | Speed | Notes |
|---|---|---|---|
| **GPU 4-bit** | ~5 GB | 30-50 tok/s | RTX 3060+ / T4+ / RTX 5070 |
| **GPU 8-bit** | ~8 GB | 20-35 tok/s | Higher quality |
| **CPU** (llama.cpp) | ~6 GB RAM | 5-15 tok/s | No GPU needed |
| **Apple Silicon** | ~5 GB | 20-40 tok/s | M1/M2/M3/M4 |

GGUF: **0.0 GB** (Q4_K_M)

---

## Provenance

| | |
|---|---|
| **Base** | [`Qwen2.5-7B-Instruct`](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) via `unsloth/Qwen2.5-7B-Instruct-bnb-4bit` |
| **Method** | LoRA SFT (r=16, α=32, dropout=0) |
| **Dataset** | [`zkaedi/zkaedi-prime-constitutions`](https://huggingface.co/datasets/zkaedi/zkaedi-prime-constitutions) — 127 pairs |
| **Teacher** | Qwen2.5-72B-Instruct-AWQ |
| **Steps** | 60 burn-in, batch 8 (2×4), lr 2e-4 linear |
| **Quantization** | bf16 → Q4_K_M |
| **Verification** | Ouroboros 19/19 ✅ |

> _Loss data not provided. Use `--loss-json outputs/trainer_state.json`_

### Inference Benchmark

| Metric | Value |
|---|---|
| GPU | NVIDIA A100-SXM4-80GB (79.3GB) |
| Model load | 31.7s |
| Input tokens | 3072 |
| Generated | 1024 tokens in 50.0s |
| Throughput | 20.5 tok/s |

---

## Files

| File | Size |
|---|---|
| `Qwen2.5-7B-Instruct.Q4_K_M.gguf` | 0.0 MB |
| `lora_adapters/` | — |

---

## Usage

### Ollama
```bash
cat > Modelfile << 'EOF'
FROM ./Qwen2.5-7B-Instruct.Q4_K_M.gguf
PARAMETER temperature 0.7
PARAMETER top_p 0.9
SYSTEM You are ZKAEDI PRIME. Analyze Hamiltonian field traces and output constitutional analysis.
EOF
ollama create zkaedi-mini -f Modelfile
ollama run zkaedi-mini

llama.cpp

./llama-cli -m Qwen2.5-7B-Instruct.Q4_K_M.gguf \
  --temp 0.7 --top-p 0.9 -n 2048 \
  -p "You are ZKAEDI PRIME. Analyze the Hamiltonian field trace below and output the highly accurate, mathematical Constitution v2.

### Hamiltonian Field Trace:
<your trace>

### Constitution v2:"

Python (transformers + PEFT)

from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel
import torch

bnb = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_quant_type="nf4",
                         bnb_4bit_compute_dtype=torch.bfloat16)
base = "Qwen/Qwen2.5-7B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base, quantization_config=bnb, device_map="auto")
model = PeftModel.from_pretrained(model, "zkaedi/ZKAEDI-MINI-GGUF", subfolder="lora_adapters")
model = model.merge_and_unload()
model.eval()

inputs = tokenizer("Your prompt", return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=2048, temperature=0.7)
print(tokenizer.decode(out[0], skip_special_tokens=True))

Load Training Data

from datasets import load_dataset
ds = load_dataset("zkaedi/zkaedi-prime-constitutions", split="train")
# Columns: run_id, field_trace, constitution_v1, critique, constitution_v2,
#          tier, eta, gamma, beta, sigma, seed, phase_stats, sd_min/max/range

Prompt Format

You are ZKAEDI PRIME. Analyze the Hamiltonian field trace below and output the highly accurate, mathematical Constitution v2.

### Hamiltonian Field Trace:
{trace}

### Constitution v2:
{model generates}

ZKAEDI Ecosystem


🔱 ZKAEDI-MINI	← You are here
🔍 gemma-7b-solidity-energy-signatures	Vulnerability energy mapper
🛡️ solidity-vuln-auditor-7b	Audit reports
🧬 solidity-prime-v2-merged	Base (F1=0.6816)
🐙 Kraken v2.2	Bytecode (F1=1.0)
🐉 Leviathan v2	EVM Hilbert manifold
🐍 Hydra v1	Control flow
🦁 Chimera v1	Storage (F1=1.0)
🐍 Basilisk v1	External calls (F1=1.0)
🏟️ prime-swarm-hunter	Fusion voting
📊 zkaedi/zkaedi-prime-constitutions	127 trace pairs
⚡ solidity-vulnerability-energy-signatures	2250 vuln samples

Limitations

60-step burn-in (not fully converged) · 127 examples (small) · SFT only (no DPO/RLHF) · Gated model

Unsloth · Ouroboros Protocol · 🔱 ZKAEDI PRIME

Downloads last month: -

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for zkaedi/ZKAEDI-MINI-GGUF

Base model

Qwen/Qwen2.5-7B

Finetuned

Qwen/Qwen2.5-7B-Instruct

Finetuned

unsloth/Qwen2.5-7B-Instruct

Adapter

(662)

this model

zkaedi
/

ZKAEDI-MINI-GGUF

You need to agree to share your contact information to access this model

🔱 ZKAEDI-MINI-GGUF

Example

ZKAEDI PRIME CONSTITUTION - run_010_20260312_210646 Model : Qwen2.5-72B-Instruct-AWQ (H100) Matrix : eta=0.149, gamma=0.282, beta=0.237, sigma=0.104, seed=59542

Preamble

Article I: Fundamental Principles

llama.cpp

Python (transformers + PEFT)

Load Training Data

Prompt Format

ZKAEDI Ecosystem

Limitations

Model tree for zkaedi/ZKAEDI-MINI-GGUF

Dataset used to train zkaedi/ZKAEDI-MINI-GGUF