Instructions to use Nanbeige/Nanbeige4.2-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Nanbeige/Nanbeige4.2-3B with Transformers:

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

pipe = pipeline("text-generation", model="Nanbeige/Nanbeige4.2-3B", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("Nanbeige/Nanbeige4.2-3B", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Nanbeige/Nanbeige4.2-3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Nanbeige/Nanbeige4.2-3B"
# 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/Nanbeige4.2-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Nanbeige/Nanbeige4.2-3B

SGLang

How to use Nanbeige/Nanbeige4.2-3B 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/Nanbeige4.2-3B" \
    --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/Nanbeige4.2-3B",
		"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/Nanbeige4.2-3B" \
        --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/Nanbeige4.2-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Nanbeige/Nanbeige4.2-3B with Docker Model Runner:
```
docker model run hf.co/Nanbeige/Nanbeige4.2-3B
```

Suitability for 8 GB graphics cards

#18

by thermi6 - opened about 21 hours ago

Discussion

thermi6

about 21 hours ago

Thank you so much for your excellent model! It is very useful and usable for small contexts on 8 GB graphics cards. However, as stated in a now closed discussion, the KV cache increases in size very quickly. Then it is not possible anymore to keep the data in VRAM.

Would it be possible to aim at severly limiting the size increase of the KV cache in the near future?

Right now this works on 8 GB cards: 4 bit quant + 32768 tokens context length. Any higher and it is too large for VRAM.

I think you would agree that just 32768 tokens is a bit small. What is your opinion on the matter? Do you have any pointers for improving the situation with the current model?

leran1995

Nanbeige LLM Lab org about 17 hours ago

Hi, thank you for the kind words and for the detailed report!
Both improving performance and reducing KV cache are on our roadmap. As mentioned in the earlier discussion, we did try sharing the KV cache across loop passes, but it noticeably hurt performance, so we kept the full cache in Nanbeige4.2. Going forward, we are exploring Linear / Sparse Attention to bring the long-context memory cost down more fundamentally. Stay tuned!

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