You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
感谢🙏from 敖武在 2023年6月16日,16:42,xiningnlp ***@***.***> 写道:
需要尝试一下指令微调。不过ChatGLM-6B迄今为止没有公布他们如何将GLM-XXX微调为ChatGLM-6B的。
—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>
ChatGLM-6B对外暴露的是对话式API,GLM-10B暴露的是文本补全式的API,如何能将GLM-10B封装成对话式API呢?
我尝试过将对话内容封装成对话文本让GLM-10B进行补全,但是效果很不好,求建议❤️❤️❤️
原理:将对话内容封装成一段字符串,每次到AI回答的时候就将历史聊天记录封装成字符串,由AI进行补全
封装成对话文本如下:
对话效果如下:
对话效果很差,AI机器人一直在重复。
The text was updated successfully, but these errors were encountered: