Instructions to use stamsam/maple-preview-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use stamsam/maple-preview-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 stamsam/maple-preview-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf stamsam/maple-preview-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 stamsam/maple-preview-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf stamsam/maple-preview-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 stamsam/maple-preview-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf stamsam/maple-preview-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 stamsam/maple-preview-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf stamsam/maple-preview-gguf:Q4_K_M

Use Docker

docker model run hf.co/stamsam/maple-preview-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use stamsam/maple-preview-gguf with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/stamsam/maple-preview-gguf:Q4_K_M

Ollama
How to use stamsam/maple-preview-gguf with Ollama:
```
ollama run hf.co/stamsam/maple-preview-gguf:Q4_K_M
```

Unsloth Studio

How to use stamsam/maple-preview-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 stamsam/maple-preview-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 stamsam/maple-preview-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for stamsam/maple-preview-gguf to start chatting

How to use stamsam/maple-preview-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf stamsam/maple-preview-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": "stamsam/maple-preview-gguf:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

OpenClaw new

How to use stamsam/maple-preview-gguf with OpenClaw:

Start the llama.cpp server

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

Lemonade

How to use stamsam/maple-preview-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull stamsam/maple-preview-gguf:Q4_K_M

Run and chat with the model

lemonade run user.maple-preview-gguf-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use stamsam/maple-preview-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 stamsam/maple-preview-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 stamsam/maple-preview-gguf:Q4_K_M

Run Hermes

hermes

Atomic Chat

CUDA TQ2_0 fixes: 7.33x pp512 speedup with reproducible evidence

by x0me - opened 3 days ago

Discussion

x0me

3 days ago

Hugging Face CLI

I reproduced the Maple TQ2_0 GPU anomaly in the llama.cpp fork and found that it was primarily a CUDA routing/kernel problem, not evidence that ternary weights inherently favor CPUs.

The patch series makes the existing batch-one MMVQ route reachable, vectorizes the ternary dot product, adds CPU-reference coverage, and implements/enables a fused TQ2_0 MMQ prompt path.

Fresh controlled RTX 4080 SUPER validation (CUDA 12.8, driver 610.62):

pp512: 1,457 → 10,674 t/s (7.33×) with MMQ enabled
tg128 median: 375.0 → 374.8 t/s (effectively unchanged, as expected)
correctness: 31/31 MUL_MAT + 72/72 MUL_MAT_ID against the CPU backend

The enabled path also passed all 103 tests on sm_86 and sm_120. Harvested unverified Vast results were 6,985.69 pp512 / 342.73 tg128 on RTX 3080 (sm_86) and 5,738.43 / 323.89 on RTX 5060 Ti (sm_120).

Reproducible dataset, raw samples, hashes, exact patches, and harness:
https://huggingface.co/datasets/x0me/maple-preview-cuda-benchmarks

Implementation repository:
https://github.com/PascalAI2024/maple-preview-windows-cuda

Important limits: I have not verified A100/sm_80 yet; the Vast hosts are explicitly marked unverified; and this validates throughput/correctness, not the model's coding, multilingual, or instruction-following quality.

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