[Bug]:使用qwen-vl-plus模型报错400错误
haijian-wang opened this issue · comments
Describe` the bug
当我在模型配置中使用了dashscope api的识图模型qwen-vl-plus,在使用Msg模块传入图片给agent时报错400,url换过本地和网络,都是报错400。
To Reproduce
Steps to reproduce the behavior:
- You code
import agentscope
from agentscope.agents import DialogAgent,UserAgent
from agentscope.message import Msg
from agentscope.pipelines.functional import sequentialpipeline
agentscope.init(model_configs=[
{
"config_name": "qwen-vl-plus",
"model_type": "dashscope_text_embedding",
"model_name": "qwen-vl-plus",
"api_key": "sk-df---314ae----------------------"
}
])
agent_1 = DialogAgent(
name="助手",
sys_prompt="你是一位好帮手",
model_config_name="qwen-vl-plus"
)
useragent = UserAgent()
x = Msg(name='用户',content='帮我描述一下这个图片',url="https://dashscope.oss-cn-beijing.aliyuncs.com/images/dog_and_girl.jpeg")
print(x)
while x is None or x.content != "exit":
x = sequentialpipeline([agent_1,useragent])
- How to execute
我是从本地安装的
git clone https://github.com/modelscope/agentscope.git
cd AgentScope
pip install -e .
- See error
Traceback (most recent call last):
File "D:\python_project_yeah\NLP\agent\7agent处理图像测试.py", line 23, in <module>
x = sequentialpipeline([agent_1,useragent])
File "D:\python_project_yeah\NLP\agentscope\src\agentscope\pipelines\functional.py", line 44, in sequentialpipeline
msg = operators[0](x)
File "D:\python_project_yeah\NLP\agentscope\src\agentscope\agents\agent.py", line 117, in __call__
res = self.reply(*args, **kwargs)
File "D:\python_project_yeah\NLP\agentscope\src\agentscope\agents\dialog_agent.py", line 81, in reply
response = self.model(prompt).text
File "D:\python_project_yeah\NLP\agentscope\src\agentscope\models\model.py", line 166, in checking_wrapper
return model_call(self, *args, **kwargs)
File "D:\python_project_yeah\NLP\agentscope\src\agentscope\models\dashscope_model.py", line 412, in __call__
raise RuntimeError(error_msg)
RuntimeError: Request id: 4b565370-9acd-9eda-8a6c-94668a1fe8b1, Status code: 400, error code: InvalidParameter, error message: url error, please check url!.
Expected behavior
成功运行
Environment (please complete the following information):
- AgentScope Version:0.0.1
- Python Version: 3.9
- OS: windows
There seems to be a misunderstanding regarding the functionality of our current implementation of the DashScope wrapper. At present, it includes capabilities for chat
, image synthesis
, and text embedding
but does not extend to multimodal interactions. The error you've encountered stems from the specified "model_type": "dashscope_text_embedding"
within your model_configs
. This configuration explicitly invokes the dashscope.TextEmbedding.call
method, which unfortunately does not support the "model_name": "qwen-vl-plus"
as indicated by the URL error message.
If your intention is to utilize dashscope.MultiModalConversation.call
, I shall inform you that this interface is not yet implemented in our system. Maybe we will implement it later. However, if you're keen on accessing this functionality, you are welcome to develop a DashScopeMultiModalWrapper
by following the existing implementation patterns found in the dashscope_model.py
. For reference on the parameters required, please visit https://help.aliyun.com/zh/dashscope/developer-reference/tongyi-qianwen-vl-plus-api.
We would also be delighted to consider any pull requests to our library should you decide to contribute your implementation.
Currently, we do not have support for multimodal reasoning functionality, but we plan to continuously extend our support for additional models.
If multimodal reasoning capabilities are essential for your needs, we highly encourage you to take the initiative to implement a DashScopeMultiModalWrapper
. For guidance, you can refer to the implementation of the other wrappers found within dashscope_model.py
. Your contributions are valuable, and we warmly invite you to submit pull requests to our repository.
好的谢谢,我会去试试的,感谢你的解答!