mirror of
https://github.com/deepseek-ai/DeepSeek-R1.git
synced 2026-11-05 00:47:42 +00:00
[GH-ISSUE #113] 关于 DeepSeek-R1 蒸馏过程的疑问 #95
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/DeepSeek-R1#95
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @Alphayellowcat on GitHub (Jan 28, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/113
描述
您好,感谢 DeepSeek 团队开源这么强大的模型!我在阅读 DeepSeek-R1 的论文时,对其中的蒸馏过程产生了一些疑问,希望能获得更多详细信息。
问题背景
在论文的蒸馏部分,提到:
同时,实验结果表明,蒸馏出的 Qwen-32B 在推理任务上明显优于直接用 RL 训练的版本。
然而,论文并未详细说明蒸馏的具体过程,例如:
我的疑问
期待的帮助
如果可以的话,希望团队能够提供:
再次感谢 DeepSeek 团队的工作,希望能收到您的解答!🙏
@xidongyuandong commented on GitHub (Feb 5, 2025):
paper提到的就是 sft 得到的蒸馏数据
@Syno8 commented on GitHub (Feb 5, 2025):
话说你们有测试,这样蒸馏后的模型,除了推理能力,对于通用任务的能力提升是如何的嘛?
@QuantaoYao commented on GitHub (Feb 8, 2025):
I am also confused by this question. I would greatly appreciate it if the Deepseek team could provide some clarification. Thank you in advance for your response
@xpdd123 commented on GitHub (Feb 10, 2025):
看了论文中的描述,也看了girhub开源的复现代码,感觉只是用DeepSeek-R1的数据,因此问题1和3不存在,论文中提到蒸馏用到了推理数据和非推理数据,因此有COT数据,训练过程loss和普通的LLM训练应该一样
@CzsGit commented on GitHub (Feb 10, 2025):
Nevertheless, he introduced novel tokens and made modifications to the prompt concatenation format during inference. These alterations suggest that the process deviates from standard LLM distillation practices, or certain details have not been disclosed.
@GuoYi0 commented on GitHub (Feb 11, 2025):
我使用llama_factory库,对Qwen2.5-Math-1.5B进行蒸馏,使用的数据集在这里。发现总是有很多幻想。求大神指教。有兴趣的老哥可以加我微信,一起学习 guoyi19930
@haichuan1221 commented on GitHub (Feb 13, 2025):
模型大于3B才有推理能力
@zhufeizzz commented on GitHub (Feb 13, 2025):
其实我也有疑问,根据论文描述是80w数据集直接进行SFT,并没有用到Token概率分布的软标签。这算传统意义上的蒸馏吗?
@wingjson commented on GitHub (Feb 13, 2025):
该啥样啥样,推理强了
@freedom-cui commented on GitHub (Feb 23, 2025):
Hello 我想和大家请教一个问题DeepSeek R1的蒸馏,到底是teacher-student这种传统意义上的蒸馏,还是说指用一个较大模型(或者是具有推理能力模型)推理的数据进行训练的过程
@rainfallLLF commented on GitHub (Mar 7, 2025):
后者,所谓的“知识蒸馏”
@github-actions[bot] commented on GitHub (Apr 9, 2025):
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you believe this issue is still relevant, please leave a comment to keep it open. Thank you for your contributions!
@github-actions[bot] commented on GitHub (May 9, 2025):
false