Instructions to use lmdeploy/llama2-chat-70b-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lmdeploy/llama2-chat-70b-4bit with Transformers:

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

pipe = pipeline("text-generation", model="lmdeploy/llama2-chat-70b-4bit")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("lmdeploy/llama2-chat-70b-4bit")
model = AutoModelForCausalLM.from_pretrained("lmdeploy/llama2-chat-70b-4bit")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use lmdeploy/llama2-chat-70b-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lmdeploy/llama2-chat-70b-4bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lmdeploy/llama2-chat-70b-4bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/lmdeploy/llama2-chat-70b-4bit

SGLang

How to use lmdeploy/llama2-chat-70b-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 "lmdeploy/llama2-chat-70b-4bit" \
    --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": "lmdeploy/llama2-chat-70b-4bit",
		"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 "lmdeploy/llama2-chat-70b-4bit" \
        --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": "lmdeploy/llama2-chat-70b-4bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use lmdeploy/llama2-chat-70b-4bit with Docker Model Runner:
```
docker model run hf.co/lmdeploy/llama2-chat-70b-4bit
```

llama2-chat-70b-w4 weights

by sandrobovelli - opened Aug 25, 2023

Discussion

sandrobovelli

Aug 25, 2023

•

edited Aug 25, 2023

Hi. I would like to know when the 4bit quantized 70B weights will be released.
I'm trying to quantize the original hf checkpoint, but the script constantly run out of memory (my system RAM is around 96GB).
What is the approximate memory requirement to quantize the 70B model?

viktor-ferenczi

Sep 8, 2023

Add more swap and run again. But still, having the weights here would be nice.

unsubscribe

lmdeploy org Nov 14, 2023

@sandrobovelli @viktor-ferenczi updated

unsubscribe changed discussion status to closed Nov 14, 2023

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