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
---- 回复的原邮件 ***@***.***>发送日期2023年05月11日 09:09 ***@***.***> ***@***.***>主题[SunnyGJing/t5-pegasus-chinese] 文本摘要 (Issue #25)
你好,我想请问一下,微调时我的数据实验结果不太理想,如果我想给模型喂一些先验知识,让他能够学习到这些,用先验知识辅助完成摘要的话,是需要在预训练原模型上进行更改重新训练,还是在微调层面对输入结构进行更改就可以呢?
—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>
[
{
***@***.***": "http://schema.org",
***@***.***": "EmailMessage",
"potentialAction": {
***@***.***": "ViewAction",
"target": "#25",
"url": "#25",
"name": "View Issue"
},
"description": "View this Issue on GitHub",
"publisher": {
***@***.***": "Organization",
"name": "GitHub",
"url": "https://github.com"
}
}
]
你好,我想请问一下,微调时我的数据实验结果不太理想,如果我想给模型喂一些先验知识,让他能够学习到这些,用先验知识辅助完成摘要的话,是需要在预训练原模型上进行更改重新训练,还是在微调层面对输入结构进行更改就可以呢?
The text was updated successfully, but these errors were encountered: