Instructions to use vcruz305/Hy3-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vcruz305/Hy3-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="vcruz305/Hy3-GGUF",
	filename="Hy3-IQ2_M/Hy3-IQ2_M-00001-of-00003.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use vcruz305/Hy3-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 vcruz305/Hy3-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf vcruz305/Hy3-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 vcruz305/Hy3-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf vcruz305/Hy3-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 vcruz305/Hy3-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf vcruz305/Hy3-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 vcruz305/Hy3-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf vcruz305/Hy3-GGUF:Q4_K_M

Use Docker

docker model run hf.co/vcruz305/Hy3-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use vcruz305/Hy3-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vcruz305/Hy3-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": "vcruz305/Hy3-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/vcruz305/Hy3-GGUF:Q4_K_M

Ollama
How to use vcruz305/Hy3-GGUF with Ollama:
```
ollama run hf.co/vcruz305/Hy3-GGUF:Q4_K_M
```

Unsloth Studio

How to use vcruz305/Hy3-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 vcruz305/Hy3-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 vcruz305/Hy3-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for vcruz305/Hy3-GGUF to start chatting

How to use vcruz305/Hy3-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf vcruz305/Hy3-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": "vcruz305/Hy3-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use vcruz305/Hy3-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 vcruz305/Hy3-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 vcruz305/Hy3-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use vcruz305/Hy3-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf vcruz305/Hy3-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 "vcruz305/Hy3-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 vcruz305/Hy3-GGUF with Docker Model Runner:
```
docker model run hf.co/vcruz305/Hy3-GGUF:Q4_K_M
```

Lemonade

How to use vcruz305/Hy3-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull vcruz305/Hy3-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Hy3-GGUF-Q4_K_M

List all available models

lemonade list

Hot stuff. Llama.cpp?

by BingoBird - opened 1 day ago

Discussion

BingoBird

1 day ago

•

edited 1 day ago

Would be amazing if the PR worked already.

ETA many hours for me. Anyone have experiemce?

vcruz305

Owner 9 minutes ago

Yes — works today, with one important detail on which PR.

The PR this repo was originally converted with (#25364) is now closed; the live one is [PR #25395](https://github.com/ggml-org/llama.cpp/pull/25395), which both loads these GGUFs and adds working MTP speculative decoding. Build:

```bash
git clone https://github.com/ggml-org/llama.cpp
cd llama.cpp
git fetch origin pull/25395/head:hy3-mtp && git checkout hy3-mtp
cmake -B build -DGGML_CUDA=ON -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release -j"$(nproc)"
```

Verified on my hardware (DGX Spark GB10, single node) with IQ2_M:

- 18.0 tok/s baseline → 22.8 tok/s with MTP spec decode (+27%), 90% draft acceptance
- Flags: `--spec-type draft-mtp --spec-draft-n-max 2 --spec-draft-p-min 0.75 --parallel 1` — the `p_min 0.75` is load-bearing (default p_min makes speculation slower on this model), and `--parallel 1` is required for draft-mtp
- `--jinja` is required or the chat template aborts

All quants in this repo were re-uploaded on 07-07 with the `hy_v3` arch string PR #25395 expects, so a fresh download works out of the box. Since your download is still ahead of you: check the fit table in the README first and grab the tier that matches your VRAM — and if you only fetched some shards already, make sure your `-00001-of-*` shard is from after 07-07.

Known rough edges before you sink hours in: native OpenAI-style tool calling 500s (llama.cpp has no parser for Hy3's bespoke tool format yet — prompt-injected tools work fine), and there's a tokenizer EOG quirk that can leak `<｜hy_eos:opensource｜>` / occasionally loop — an explicit stop string on that token covers it. Details in discussion #2.

First load of a 100GB+ quant takes ~7-8 min before generation starts — don't kill it early. 🙂

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment