Instructions to use Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit with Transformers:

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

pipe = pipeline("text-generation", model="Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit")
model = AutoModelForCausalLM.from_pretrained("Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit")
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

vLLM

How to use Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit

SGLang

How to use Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit 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 "Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit" \
    --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": "Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit",
		"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 "Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit" \
        --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": "Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit with Docker Model Runner:
```
docker model run hf.co/Efficient-ML/LLaMA-3-8B-SmoothQuant-4bit-4bit
```

Why is this exactly the same size as the 8-bit one?

by dagelf - opened Apr 30, 2024

Discussion

dagelf

Apr 30, 2024

I'm guessing there is a mistake...?

Xingyu-Zheng

Efficient Intelligence and Systems org Apr 30, 2024

In our research, to quickly validate the effectiveness of various quantization methods, we only performed fake-quant on smoothquant without storing in real 4-bit format. Therefore, the size of the checkpoints we saved and uploaded is actually equivalent to the fp16 model.🪄
We will continue to improve our work to achieve as realistic quantization testing as possible with software and hardware support. More work is on the way!🤗

dagelf

Apr 30, 2024

That makes sense, I guess I could've looked :-D Thank you for the clarification!

dagelf changed discussion status to closed Apr 30, 2024

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