Instructions to use kousw/bitnet_b1_58-3B_quantized with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kousw/bitnet_b1_58-3B_quantized with Transformers:

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

pipe = pipeline("text-generation", model="kousw/bitnet_b1_58-3B_quantized")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("kousw/bitnet_b1_58-3B_quantized")
model = AutoModelForCausalLM.from_pretrained("kousw/bitnet_b1_58-3B_quantized", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use kousw/bitnet_b1_58-3B_quantized with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kousw/bitnet_b1_58-3B_quantized"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kousw/bitnet_b1_58-3B_quantized",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/kousw/bitnet_b1_58-3B_quantized

SGLang

How to use kousw/bitnet_b1_58-3B_quantized 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 "kousw/bitnet_b1_58-3B_quantized" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kousw/bitnet_b1_58-3B_quantized",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "kousw/bitnet_b1_58-3B_quantized" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kousw/bitnet_b1_58-3B_quantized",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use kousw/bitnet_b1_58-3B_quantized with Docker Model Runner:
```
docker model run hf.co/kousw/bitnet_b1_58-3B_quantized
```

bitnet_b1_58-3B_quantized / quantization.py

kousw

Upload 21 files

29964ce verified over 2 years ago

Raw

History Blame Contribute Delete

886 Bytes

	import argparse

	import torch

	from modeling_bitnet import BitnetForCausalLM
	from tokenization_bitnet import BitnetTokenizer

	torch.set_grad_enabled(False)


	parser = argparse.ArgumentParser()
	parser.add_argument("--hf_path", default="1bitLLM/bitnet_b1_58-3B", type=str)
	parser.add_argument("--output_path", default="./bitnet_b1_58-3B_quantized", type=str)


	def main(args):
	model = BitnetForCausalLM.from_pretrained(
	args.hf_path,
	device_map="auto",
	low_cpu_mem_usage=True,
	use_flash_attention_2=True,
	torch_dtype=torch.float16,
	).half()
	tokenizer = BitnetTokenizer.from_pretrained(args.hf_path, use_fast=False)

	model.quantize()

	model.save_pretrained(args.output_path, max_shard_size="5GB")

	print("Quantized model saved to", args.output_path)


	if __name__ == "__main__":
	args = parser.parse_args()
	main(args)