Instructions to use Qwen/Qwen2-72B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Qwen/Qwen2-72B with Transformers:

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

pipe = pipeline("text-generation", model="Qwen/Qwen2-72B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-72B")
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-72B")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Qwen/Qwen2-72B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Qwen/Qwen2-72B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Qwen/Qwen2-72B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Qwen/Qwen2-72B

SGLang

How to use Qwen/Qwen2-72B 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 "Qwen/Qwen2-72B" \
    --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": "Qwen/Qwen2-72B",
		"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 "Qwen/Qwen2-72B" \
        --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": "Qwen/Qwen2-72B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Qwen/Qwen2-72B with Docker Model Runner:
```
docker model run hf.co/Qwen/Qwen2-72B
```

推理速度慢，求加速推理指导

by Alanxc - opened Aug 16, 2024

Discussion

Alanxc

Aug 16, 2024

目前我是用ollama拉取的qwen2 72b，langchain设计任务链，然后用flask封装了一个接口（post方法），然后用postman测得，我的业务任务是文本分析生成JSON数据，但是现在处理一次的速度慢的可怕，请求朋友们指导一下这个要怎么加速比较合理啊。我硬件设备用的是H800 80GB *2 pcie，但是只是推理的话，72b也跑不满一张卡，应该不是硬件的问题吧。

Any recommend all welcome

Alanxc

Aug 16, 2024

reasoning speed is quite slow of my target task, here is my work flow, i used the OLLAMA to pull the qwen2:72b，and then use the langchain to build a work chain, and give prompt bulabula,finally i used the flask(python web lib) to build a service api provide for the service, when i use postman to test it , the reasoning speed are so slow, and my hard ware set is H800 80GB * 2， one thing i can believe is it's might not the hardware problem, cause when the llm work the Video memory only used 40GB+, the question is how can i make the reasoning speed upupupupupupupupupup,pls. any help are welcome ~

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