Instructions to use bunnycore/QwQen-3B-LCoT-R1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use bunnycore/QwQen-3B-LCoT-R1 with Transformers:

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

pipe = pipeline("text-generation", model="bunnycore/QwQen-3B-LCoT-R1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("bunnycore/QwQen-3B-LCoT-R1")
model = AutoModelForCausalLM.from_pretrained("bunnycore/QwQen-3B-LCoT-R1")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use bunnycore/QwQen-3B-LCoT-R1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "bunnycore/QwQen-3B-LCoT-R1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "bunnycore/QwQen-3B-LCoT-R1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/bunnycore/QwQen-3B-LCoT-R1

SGLang

How to use bunnycore/QwQen-3B-LCoT-R1 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 "bunnycore/QwQen-3B-LCoT-R1" \
    --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": "bunnycore/QwQen-3B-LCoT-R1",
		"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 "bunnycore/QwQen-3B-LCoT-R1" \
        --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": "bunnycore/QwQen-3B-LCoT-R1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use bunnycore/QwQen-3B-LCoT-R1 with Docker Model Runner:
```
docker model run hf.co/bunnycore/QwQen-3B-LCoT-R1
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

When using the QwQen-3B-LCoT-R1 model, you might notice that it can sometimes produce repetitive outputs, especially in certain contexts or with specific prompts. This is a common behavior in language models, but don’t worry—it can be managed effectively by tweaking the model’s repetition parameters.

To reduce repetition, you can experiment with the following settings:

Repetition Penalty: This parameter discourages the model from repeating the same words or phrases by applying a penalty. A higher value (e.g., 1.0) will push the model to avoid repetition more aggressively.
Temperature: This controls the randomness of the output. Lowering the temperature (e.g., 0.7) makes the model more focused and less likely to repeat itself, though it may reduce creativity slightly.

System Prompt:

Think about the reasoning process in the mind first, then provide the answer.

The reasoning process should be wrapped within <think> </think> tags, then provide the answer after that, i.e., <think> reasoning process here </think> answer here.

Configuration

The following YAML configuration was used to produce this model:


base_model: bunnycore/QwQen-3B-LCoT+bunnycore/Qwen-2.5-3b-R1-lora_model-v.1
dtype: bfloat16
merge_method: passthrough
models:
  - model: bunnycore/QwQen-3B-LCoT+bunnycore/Qwen-2.5-3b-R1-lora_model-v.1
tokenizer_source: bunnycore/QwQen-3B-LCoT