Skip to content
Discussion options

You must be logged in to vote

“在rtx5090下,跑了预训练8g的数据集和sft1g的数据集耗时最起码需要3-天” 在你的设定下可能是

当前项目在0.2B-A0.06B的RTX 3090训10GB+16GB是30h

“训练出来的效果也并不是很好,训练中探针测试,也是效果很差。”

我没有看到任何case,也不知道你的测试方法,无法给出进一步评价

“在预训练中,有时能续写出完整的语句”

这不符合当前项目现象,建议复现仓库官方代码和数据来给结论,而不作为普遍的issue发出

Replies: 2 comments 5 replies

Comment options

You must be logged in to vote
0 replies
Answer selected by jingyaogong
Comment options

You must be logged in to vote
5 replies
@jingyaogong
Comment options

@yufarui
Comment options

@yufarui
Comment options

@jingyaogong
Comment options

@yufarui
Comment options

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Category
Q&A
Labels
None yet
2 participants
Converted from issue

This discussion was converted from issue #744 on April 22, 2026 08:15.