🤗 Hugging Face | 🤖 ModelScope | 📑 Blog | 🖥️ Demo
WeChat (微信) | Discord | API
Visit our Hugging Face or ModelScope organization (click links above), search checkpoints with names starting with Qwen2-beta-
, and you will find all you need! Enjoy!
This time, we upgrade Qwen to Qwen2-beta, the beta version of Qwen2. Similar to Qwen, it is still a decoder-only transformer model with SwiGLU activation, RoPE, multi-head attention. At this moment, we have achieved:
- 6 model sizes: 0.5B, 1.8B, 4B, 7B, 14B, and 72B;
- Significant model quality improvements in chat models;
- Strengthened multilingual capabilities in both base and chat models;
- All models support the context length of
32768
tokens; - System prompts enabled for all models, which means roleplay is possible.
- No need of
trust_remote_code
anymore.
We have not integrated GQA and mixture of SWA and full attention in this version and we will add the features in the future version.
- 2024.02.05: We released the Qwen2-beta series.
Detailed evaluation results are reported in this 📑 blog.
transformers>=4.37.0
.
Warning
For a quickstart, I advise you to use chat models. We often use base models only for post-training. Here is a simple code snippet:
from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-beta-7B-Chat", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-beta-7B-Chat")
prompt = "Give me a short introduction to large language model."
messages = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
model_inputs = tokenizer([text], return_tensors="pt").to(device)
generated_ids = model.generate(model_inputs.input_ids, max_new_tokens=512, do_sample=True)
generated_ids = [output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)]
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
For quantized models, we advise you to use the GPTQ, AWQ, and GGUF correspondents, namely Qwen-beta-7B-Chat-GPTQ
, Qwen-beta-7B-Chat-AWQ
.
We strongly advise users especially those in mainland China to use ModelScope. snapshot_download
can help you solve issues concerning downloading checkpoints.
Download our provided GGUF files or create them by yourself, and you can directly use them with the latest llama.cpp
with a one-line command:
./main -m <path-to-file> -n 512 --color -i -cml -f prompts/chat-with-qwen.txt
We are now on Ollama, and you can use pull
and run
to make things work.
ollama pull qwen2-beta
ollama run qwen2-beta
You can also add things like ::14B
to choose different models. Visit ollama.ai for more information.
Qwen2-beta has already been supported by lmstudio.ai. You can directly use LMStudio with our GGUF files.
You can directly use text-generation-webui
for creating a web UI demo. If you use GGUF, remember to install the latest wheel of llama.cpp
with the support of Qwen2-beta.
Clone llamafile
, run source install, and then create your own llamafile with the GGUF file following the guide here. You are able to run one line of command, say ./qwen.llamafile
, to create a demo.
Now, Qwen2-beta is supported by multiple inference frameworks. Here we demonstrate the usage of vLLM
and SGLang
.
We advise you to use vLLM>=0.3.0
to build OpenAI-compatible API service. Start the server with a chat model, e.g. Qwen2-beta-7B-Chat
:
python -m vllm.entrypoints.openai.api_server --model Qwen/Qwen2-beta-7B-Chat
Then use the chat API as demonstrated below:
curl http://localhost:8000/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen/Qwen2-beta-7B-Chat",
"messages": [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Tell me something about large language models."}
]
}'
from openai import OpenAI
# Set OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://localhost:8000/v1"
client = OpenAI(
api_key=openai_api_key,
base_url=openai_api_base,
)
chat_response = client.chat.completions.create(
model="Qwen/Qwen2-beta-7B-Chat",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Tell me something about large language models."},
]
)
print("Chat response:", chat_response)
Please install SGLang
from source. Similar to vLLM
, you need to launch a server and use OpenAI-compatible API service. Start the server first:
python -m sglang.launch_server --model-path Qwen/Qwen2-beta-7B-Chat --port 30000
You can use it in Python as shown below:
from sglang import function, system, user, assistant, gen, set_default_backend, RuntimeEndpoint
@function
def multi_turn_question(s, question_1, question_2):
s += system("You are a helpful assistant.")
s += user(question_1)
s += assistant(gen("answer_1", max_tokens=256))
s += user(question_2)
s += assistant(gen("answer_2", max_tokens=256))
set_default_backend(RuntimeEndpoint("http://localhost:30000"))
state = multi_turn_question.run(
question_1="What is the capital of China?",
question_2="List two local attractions.",
)
for m in state.messages():
print(m["role"], ":", m["content"])
print(state["answer_1"])
We provide a colab notebook for you to experience the simplest way of finetuning. However, we advise you to turn to more advanced training frameworks, including Axolotl, Llama-Factory, Swift, etc.
Check the license of each model inside its HF repo. It is NOT necessary for you to submit a request for commercial usage.
If you find our work helpful, feel free to give us a cite.
@article{qwen,
title={Qwen Technical Report},
author={Jinze Bai and Shuai Bai and Yunfei Chu and Zeyu Cui and Kai Dang and Xiaodong Deng and Yang Fan and Wenbin Ge and Yu Han and Fei Huang and Binyuan Hui and Luo Ji and Mei Li and Junyang Lin and Runji Lin and Dayiheng Liu and Gao Liu and Chengqiang Lu and Keming Lu and Jianxin Ma and Rui Men and Xingzhang Ren and Xuancheng Ren and Chuanqi Tan and Sinan Tan and Jianhong Tu and Peng Wang and Shijie Wang and Wei Wang and Shengguang Wu and Benfeng Xu and Jin Xu and An Yang and Hao Yang and Jian Yang and Shusheng Yang and Yang Yao and Bowen Yu and Hongyi Yuan and Zheng Yuan and Jianwei Zhang and Xingxuan Zhang and Yichang Zhang and Zhenru Zhang and Chang Zhou and Jingren Zhou and Xiaohuan Zhou and Tianhang Zhu},
journal={arXiv preprint arXiv:2309.16609},
year={2023}
}
If you are interested to leave a message to either our research team or product team, join our Discord or WeChat groups!