Instructions to use Nanbeige/Nanbeige4.1-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Nanbeige/Nanbeige4.1-3B with Transformers:

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

pipe = pipeline("text-generation", model="Nanbeige/Nanbeige4.1-3B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Nanbeige/Nanbeige4.1-3B")
model = AutoModelForCausalLM.from_pretrained("Nanbeige/Nanbeige4.1-3B")
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 Nanbeige/Nanbeige4.1-3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Nanbeige/Nanbeige4.1-3B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Nanbeige/Nanbeige4.1-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Nanbeige/Nanbeige4.1-3B

SGLang

How to use Nanbeige/Nanbeige4.1-3B 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 "Nanbeige/Nanbeige4.1-3B" \
    --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": "Nanbeige/Nanbeige4.1-3B",
		"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 "Nanbeige/Nanbeige4.1-3B" \
        --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": "Nanbeige/Nanbeige4.1-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Nanbeige/Nanbeige4.1-3B with Docker Model Runner:
```
docker model run hf.co/Nanbeige/Nanbeige4.1-3B
```

Reduce Reasoning Effort and/or Toggle Thinking to low?

#40

by mofhayat - opened Mar 10

Discussion

mofhayat

Mar 10

•

edited Mar 10

I'm trying reduce Nanbeige's reasoning efforts and reduce reasoning time for cases when it seems to get the right answer early in it's thought chain.

I am using the AutoModelForCausalLM lib with the pretrained base model (most recent version as of 3/10/26). I plan on fine tuning this model but I first wanted to set a base benchmark. I have noticed from some initial testing that I am getting correct answers early in the logic chain but then the model keeps thinking. Eventually it still gets the right answer but after much thought.

I imagine after some fine tuning it will reduce it's reasoning time for my usecase, but I wanted to first understand if there are some settings I can change to help achieve this.

I've tried looking for a reasoning_effort arg (similar to Grok) but do not see one. I also tried to remove the thinking logs from the response to see if that would help by adding/no_think ( similar to Qwen models source to the end of System messages and User messages with no luck. I've kept the temps between 0.2 and 0.7

I'm a bit new to playing around with LLMs at this level so apologies if this is a naive question.

As a reference; here are some settings I have configured

    tokenizer = AutoTokenizer.from_pretrained(
    'Nanbeige/Nanbeige4.1-3B',
    use_fast=False,
    trust_remote_code=True
    )
    
    model = AutoModelForCausalLM.from_pretrained(
    'Nanbeige/Nanbeige4.1-3B',
    torch_dtype='auto',
    trust_remote_code=True
    )
    # move model to MPS
    model = model.to("mps")

    messages = [
        {'role': 'system', 'content': SYSTEM_MESSAGE},
        {'role': 'user', 'content': user_content}
    ]
    prompt = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    tokenize=False
    )
    input_ids = tokenizer(prompt, add_special_tokens=False, return_tensors='pt').input_ids
    output_ids = model.generate(input_ids.to('mps'), max_new_tokens=1000, eos_token_id=166101, temperature=0.2, do_sample=True)
    
    resp = tokenizer.decode(output_ids[0], skip_special_tokens=True)

mofhayat

Mar 11

Seems to be a known issue: https://huggingface.co/Nanbeige/Nanbeige4.1-3B/discussions/19#6997a4b034b384dc490f2526

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