We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
~
No response
[ { "instruction": "人类指令(必填)", "input": "人类输入(选填)", "output": "模型回答(必填)", "system": "系统提示词(选填)", "history": [ ["第一轮指令(选填)", "第一轮回答(选填)"], ["第二轮指令(选填)", "第二轮回答(选填)"] ] } ]
例如以上数据集,模型只预测"模型回答(必填)",不预测第一轮第二轮回答。
还有就是在历史消息中指令为空的时候,能不能将第n轮回答直接拼到模型将要预测的部分中,这样模型就能直接预测"模型回答(必填)",就像这个 inputs: <|im_start|>system 系统提示词(选填)<|im_end|> <|im_start|>user 人类指令(必填) 人类输入(选填)<|im_end|> <|im_start|>assistant 第一轮回答(选填) 第二轮回答(选填) 模型回答(必填)<|im_end|>
这样对反思链的训练效果可能要更好~
The text was updated successfully, but these errors were encountered:
同样希望这个功能可以加入,这个功能对于很多distillation的工作也很有帮助,感谢!
Sorry, something went wrong.
+1
目前解决方案是分两次预测
779aae8
Train on the last turn only
Successfully merging a pull request may close this issue.
Reminder
System Info
~
Reproduction
~
Expected behavior
No response
Others
例如以上数据集,模型只预测"模型回答(必填)",不预测第一轮第二轮回答。
还有就是在历史消息中指令为空的时候,能不能将第n轮回答直接拼到模型将要预测的部分中,这样模型就能直接预测"模型回答(必填)",就像这个
inputs:
<|im_start|>system
系统提示词(选填)<|im_end|>
<|im_start|>user
人类指令(必填)
人类输入(选填)<|im_end|>
<|im_start|>assistant
第一轮回答(选填)
第二轮回答(选填)
模型回答(必填)<|im_end|>
这样对反思链的训练效果可能要更好~
The text was updated successfully, but these errors were encountered: