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

Libraries

How to use zhichen/Qwen-WisdomVast with Transformers:

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

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

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("zhichen/Qwen-WisdomVast")
model = AutoModelForCausalLM.from_pretrained("zhichen/Qwen-WisdomVast")
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 Settings

vLLM

How to use zhichen/Qwen-WisdomVast with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/zhichen/Qwen-WisdomVast

SGLang

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

Docker Model Runner
How to use zhichen/Qwen-WisdomVast with Docker Model Runner:
```
docker model run hf.co/zhichen/Qwen-WisdomVast
```

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Qwen-WisdomVast (千问-智瀚)

介绍

Qwen-WisdomVast是以Qwen1.5-7B为底座，使用 DORA + LORA+ 的训练方法，在100w高质量中文多轮SFT数据 + 20w英文多轮SFT数据 + 2000单轮自我认知数据训练而来的大模型，数学能力相比Qwen1.5-7B-Chat提升了5.16%，在HumanEval数据集上相比Qwen1.5-7B-Chat提升了12.8，在MBPP数据集上提升了11.6%，在BBH数据集上 提升了12.44% ，全部评测表现见下表。

Github：https://github.com/seanzhang-zhichen/Qwen-WisdomVast

评测表现

Model	MMLU	C-Eval	GSM8K	MATH	HumanEval	MBPP	BBH
Qwen1.5-7B-Chat	60.88	70.18	54.13	7.96	31.10	15.00	31.67
Qwen-WisdomVast	57.09	70.82	51.93	13.12	43.90	26.60	44.11

说明：

由于官方并未公布Qwen1.5-7B-Chat的评测表现，所以我们自己使用opencompass测试得到以上结果

Qwen-WisdomVast使用和Qwen1.5-7B-Chat一样的参数进行测试

模型下载

Model	Download
Qwen1.5-7B	🤗 HuggingFace 🤖 ModelScope
Qwen-WisdomVast-Lora	🤗 HuggingFace 🤖 ModelScope
Qwen-WisdomVast (合并好的模型)	🤗 HuggingFace 🤖 ModelScope

合并LORA模型（可跳过）

1、下载 Qwen1.5-7B

git clone https://www.modelscope.cn/qwen/Qwen1.5-7B.git

2、下载Qwen-WisdomVast-Lora

From ModelScope

git lfs install
git clone https://www.modelscope.cn/seanzhang/Qwen-WisdomVast-Lora.git

From HuggingFace

git lfs install
git clone https://huggingface.co/zhichen/Qwen-WisdomVast-Lora

3、合并模型

python merge_lora.py \
    --base_model path/to/qwen/Qwen1.5-7B \
    --lora_model path/to/lora/Qwen-WisdomVast-Lora \
    --output_dir ./Qwen-WisdomVast

下载 Qwen-WisdomVast（合并好的模型）

From ModelScope

git lfs install
git clone https://www.modelscope.cn/seanzhang/Qwen-WisdomVast.git

From HuggingFace

git lfs install
git clone https://huggingface.co/zhichen/Qwen-WisdomVast

命令行推理

python cli_demo.py  --model_path ./Qwen-WisdomVast(换成你自己的合并后的模型路径)

web 推理

python web_demo.py  --model_path ./Qwen-WisdomVast(换成你自己的合并后的模型路径)

vllm web 推理

1、使用vllm部署模型

python -m vllm.entrypoints.openai.api_server --served-model-name Qwen-WisdomVast --model ./Qwen-WisdomVast(换成你自己的合并后的模型路径)

2、在命令行执行

python vllm_web_demo.py --model Qwen-WisdomVast

复现测试结果

1、使用vllm部署openai api server

部署命令:

python -m vllm.entrypoints.openai.api_server --served-model-name Qwen-WisdomVast --model ./Qwen-WisdomVast(换成你自己的合并后的模型路径)

2、使用opencompass框架进行测试

参考：使用opencompass验证模型效果

按照以上文章修改好后，将eval_qwen_wisdomvast.py文件到 opencompass/configs文件夹下

3、执行测试脚本

python run.py configs/eval_qwen_wisdomvast.py  -w outputs/Qwen-WisdomVast

LICENSE

本项目仅可应用于研究目的，项目开发者不承担任何因使用本项目（包含但不限于数据、模型、代码等）导致的危害或损失。详细请参考免责声明。

Qwen-WisdomVast项目代码的授权协议为 The Apache License 2.0，代码可免费用做商业用途，模型权重和数据只能用于研究目的。请在产品说明中附加Qwen-WisdomVast的链接和授权协议。

Citation

如果你在研究中使用了Qwen-WisdomVast，请按如下格式引用：

@misc{Qwen-WisdomVast,
  title={Qwen-WisdomVast},
  author={Zhichen Zhang, Weihan Huang},
  year={2024},
  howpublished={\url{https://github.com/seanzhang-zhichen/Qwen-WisdomVast}},
}