Instructions to use limin-arc/llama3-8b-1.58bit-pretrained with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use limin-arc/llama3-8b-1.58bit-pretrained with Transformers:

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

pipe = pipeline("text-generation", model="limin-arc/llama3-8b-1.58bit-pretrained")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("limin-arc/llama3-8b-1.58bit-pretrained")
model = AutoModelForCausalLM.from_pretrained("limin-arc/llama3-8b-1.58bit-pretrained", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use limin-arc/llama3-8b-1.58bit-pretrained with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "limin-arc/llama3-8b-1.58bit-pretrained"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "limin-arc/llama3-8b-1.58bit-pretrained",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/limin-arc/llama3-8b-1.58bit-pretrained

SGLang

How to use limin-arc/llama3-8b-1.58bit-pretrained 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 "limin-arc/llama3-8b-1.58bit-pretrained" \
    --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": "limin-arc/llama3-8b-1.58bit-pretrained",
		"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 "limin-arc/llama3-8b-1.58bit-pretrained" \
        --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": "limin-arc/llama3-8b-1.58bit-pretrained",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use limin-arc/llama3-8b-1.58bit-pretrained with Docker Model Runner:
```
docker model run hf.co/limin-arc/llama3-8b-1.58bit-pretrained
```

needs work. it appears that the 1bit models are far less efficient when scare in terms of memory usage

Downloads last month: 11

Safetensors

Model size

9B params

Tensor type

F32

limin-arc
/

llama3-8b-1.58bit-pretrained

Dataset used to train limin-arc/llama3-8b-1.58bit-pretrained