Fix an unusual issue that occurs when using OpenAIChat for llm_math #1410
Conversation
… the code style of the "Final Answer:" in Mrkl
|
Sorry, I need to add one more point of clarification. My prompt includes a "fifth root" question, and without it, OpenAIChat works fine. Of course, whether or not the prompt is included, OpenAI always works fine. I think this will help you determine whether or not this PR needs to be merged. From my personal perspective and my experience of debugging non-English languages, I still recommend this PR. |
…angchain-ai#1410) Fix an issue that occurs when using OpenAIChat for llm_math, refer to the code style of the "Final Answer:" in Mrkl。 the reason is I found a issue when I try OpenAIChat for llm_math, when I try the question in Chinese, the model generate the format like "\n\nQuestion: What is the square of 29?\nAnswer: 841", it translate the question first , then answer. below is my snapshot: <img width="945" alt="snapshot" src="https://user-images.githubusercontent.com/82029664/222642193-10ecca77-db7b-4759-bc46-32a8f8ddc48f.png">
|
hello,老哥，我在调用Memory相关功能的时候也发现它会翻译中文为英文，不知道有没有什么好办法？ |
@1973sir 我对memory源码不算熟啊。首先memory应该不会翻译。 但是你那个会自动summary，我简单看了一眼源码，llm跑了一波生成summary的功能，这个summary的prompt都是英文写的。你只是要memory的话，用 ConversationBufferMemory试试？而且你要是纯闲聊场景，直接用conversational agent看看？ |
我是在开发一个AI陪伴机器人，用memory主要是为了存储长期记忆，用记忆图谱是为了存储对话过程中用户提供的有关自己的个性化信息。我先看看你说的conversational agent。我用的gpt-3.5-turbo模型。你在搞啥项目呀？ |
|
conversational agent貌似比较适合复现类似网页版chatgpt的应用，不太适合做长期的AI陪伴机器人。ConversationBufferMemory的话只适合当前session，没法保存长期记忆。实在不行也许只能找个支持中文的summary/embedding的三方库，自己搞记忆检索/知识库存储/提取相关的组件了。 |
|
刚又看了下文档，他们新增了Chat Vector DB那些，不过我估计不支持中文。 |
|
@1973sir 这儿不适合讨论这些。你有其他联系方式吗？我们下来沟通 |
方便的话加我微信吧 pinball1973 |
也遇到了类似问题，想讨论 |
可以加微信聊哈 |
Fix an issue that occurs when using OpenAIChat for llm_math, refer to the code style of the "Final Answer:" in Mrkl。 the reason is I found a issue when I try OpenAIChat for llm_math, when I try the question in Chinese, the model generate the format like "\n\nQuestion: What is the square of 29?\nAnswer: 841", it translate the question first , then answer. below is my snapshot:
