Instructions to use microsoft/bitnet-b1.58-2B-4T with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use microsoft/bitnet-b1.58-2B-4T with Transformers:

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

pipe = pipeline("text-generation", model="microsoft/bitnet-b1.58-2B-4T", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("microsoft/bitnet-b1.58-2B-4T", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("microsoft/bitnet-b1.58-2B-4T", trust_remote_code=True)
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use microsoft/bitnet-b1.58-2B-4T with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "microsoft/bitnet-b1.58-2B-4T"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "microsoft/bitnet-b1.58-2B-4T",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/microsoft/bitnet-b1.58-2B-4T

SGLang

How to use microsoft/bitnet-b1.58-2B-4T 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 "microsoft/bitnet-b1.58-2B-4T" \
    --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": "microsoft/bitnet-b1.58-2B-4T",
		"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 "microsoft/bitnet-b1.58-2B-4T" \
        --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": "microsoft/bitnet-b1.58-2B-4T",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use microsoft/bitnet-b1.58-2B-4T with Docker Model Runner:
```
docker model run hf.co/microsoft/bitnet-b1.58-2B-4T
```

More training details?

#15

by elepedus - opened Apr 18, 2025

Discussion

elepedus

Apr 18, 2025

Hi,

Thanks so much for the model and very well-written & approachable technical report – It's great to see continuing work on BitNet!

I wonder if you could share any more details about the training, especially regarding cost / resource utilisation and how it compares to un-quantised training runs? Naively, I would expect meaningful efficiency gains at training time as well, but it would be great to get some concrete numbers.

Thanks in advance,
Ed

sd983527

Apr 24, 2025

I don't think there is any benefit in training efficiency since it still use full precision in training stage.

elepedus

Apr 24, 2025

•

edited Apr 24, 2025

I don't think there is any benefit in training efficiency since it still use full precision in training stage.

Really??

I thought the whole point of BitNet was that the ternary “quantisation” is applied during training, rather than afterwards. Otherwise why even bother training a new model in the first place?

The authors could have just applied post-training ternary quantisation to an existing model like Phi-4. It probably would have been quicker, cheaper and served to more clearly indicate that this is “just” post-training quantisation, providing a clearer, more direct baseline to compare against.

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