Instructions to use SC117/QwenPaw-Flash-9B-heretic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SC117/QwenPaw-Flash-9B-heretic with Transformers:

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

pipe = pipeline("text-generation", model="SC117/QwenPaw-Flash-9B-heretic")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("SC117/QwenPaw-Flash-9B-heretic")
model = AutoModelForMultimodalLM.from_pretrained("SC117/QwenPaw-Flash-9B-heretic", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use SC117/QwenPaw-Flash-9B-heretic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "SC117/QwenPaw-Flash-9B-heretic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SC117/QwenPaw-Flash-9B-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/SC117/QwenPaw-Flash-9B-heretic

SGLang

How to use SC117/QwenPaw-Flash-9B-heretic 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 "SC117/QwenPaw-Flash-9B-heretic" \
    --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": "SC117/QwenPaw-Flash-9B-heretic",
		"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 "SC117/QwenPaw-Flash-9B-heretic" \
        --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": "SC117/QwenPaw-Flash-9B-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use SC117/QwenPaw-Flash-9B-heretic with Docker Model Runner:
```
docker model run hf.co/SC117/QwenPaw-Flash-9B-heretic
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

GGUF Heretic

QwenPaw-Flash-9B-heretic

English | 📖 中文文档

📌 Overview

F32 safetensors of **QwenPaw-Flash-9B-heretic**, a 9B dense model fine-tuned with Heretic methodology on Qwen3.5-9B.

🧠 Model Details

- **Base model**: Qwen3.5-9B - **Precision**: F32 (float32 safetensors) - **Parameters**: ~9B - **Shards**: model-00001 ~ model-00008 (8 files, F32 main weights) - **Additional**: model-00009 (BF16, Multi-Token Prediction head extracted from Qwen3.5-9B)

⚡ MTP (Multi-Token Prediction)

`model-00009-of-00009.safetensors` contains the MTP head weights extracted from Qwen3.5-9B. MTP enables the model to predict multiple future tokens in a single forward pass, improving generation speed via speculative decoding.

MTP acceptance rate: ~43%
Speedup: ~1.5-1.9x decode throughput

For MTP-enabled GGUF inference, see the MTP GGUF repo below.

📦 GGUF Quantized Versions

For inference with llama.cpp / Ollama / LM Studio, use the GGUF versions:

Standard GGUF (no MTP): SC117/QwenPaw-Flash-9B-heretic-GGUF
MTP GGUF (with Multi-Token Prediction head): SC117/QwenPaw-Flash-9B-heretic-MTP-GGUF

🚀 Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "SC117/QwenPaw-Flash-9B-heretic",
    torch_dtype=torch.float32,
    device_map="auto",
)
tokenizer = AutoTokenizer.from_pretrained("SC117/QwenPaw-Flash-9B-heretic")

📄 License

Same as base model (Qwen3.5-9B).