Instructions to use deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B with Transformers:

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

pipe = pipeline("text-generation", model="deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B")
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B", 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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B

SGLang

How to use deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B 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 "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B" \
    --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": "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
		"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 "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B" \
        --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": "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B with Docker Model Runner:
```
docker model run hf.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
```

System Prompt

by Wanfq - opened Jan 20, 2025

Discussion

Wanfq

Jan 20, 2025

We have tested the system prompt with temperature of 0.7.

You are a helpful and harmless assistant. You should think step-by-step.

Here are the evaluation results.

Models	AIME24	MATH500	GSM8K	GPQA-Diamond	ARC-Challenge	MMLU-Pro	MMLU	LiveCodeBench
deepseek-ai/DeepSeek-R1-Distill-Qwen-32B	46.67	88.20	-	57.58	-	-	-	-

More evaluation results can be found at https://huggingface.co/FuseAI/FuseO1-DeekSeekR1-QwQ-SkyT1-32B-Preview

lewtun

Jan 22, 2025

Any idea what explains the diff in your eval scores and those from the paper?

Wanfq

Jan 22, 2025

Any idea what explains the diff in your eval scores and those from the paper?

We find the evaluation results for math and code are not correct in our current version. To address this issue, we use the code from Qwen2.5-Math and Qwen2.5-Coder for math and code evaluation. With this approach, we have successfully reproduced the results reported in the DeepSeek-R1 paper. We will update all the results tomorrow. Please stay tuned.

Wanfq

Jan 23, 2025

Any idea what explains the diff in your eval scores and those from the paper?

We have finished all the evaluation and updated the results here:

The reproduce details can be found in our blog: https://huggingface.co/blog/Wanfq/fuseo1-preview

We also provide the code in our github repo: https://github.com/fanqiwan/FuseAI/tree/main/FuseO1-Preview

Our models are in : https://huggingface.co/collections/FuseAI/fuseo1-preview-678eb56093649b2688bc9977

Have fun!

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