-
这条姑且不算是bug反馈而是算求助...因为部署流程和默认流程差别应该挺大的 参考文献:https://docs.qq.com/doc/DSkNobWtEZ2tRRk10 (安装过程中主要使用的教程,两句conda install -c没有成功,别的都做了) 平台:硬件部分: 软件部分: 模型部分: 问题描述: |
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 2 replies
-
遇到相同的问题了 :( |
Beta Was this translation helpful? Give feedback.
-
装了flashinfer,这个坎跨过去了,模型顺利进了内存 |
Beta Was this translation helpful? Give feedback.
-
我手动编译安装了这个库的whl,已经成功了
…---- 回复的原邮件 ----
| 发件人 | ***@***.***> |
| 发送日期 | 2025年04月07日 22:38 |
| 收件人 | kvcache-ai/ktransformers ***@***.***> |
| 抄送人 | waterinsect ***@***.***>,
Comment ***@***.***> |
| 主题 | Re: [kvcache-ai/ktransformers] 尝试部署Deepseek-R1 671B-Q4时出现NameError: name 'BatchMLAPagedAttentionWrapper' is not defined (Discussion #1058) |
总结:
老老实实装flashinfer
如果local_chat.py不行,试试直接用server\main.py~~,没准有奇效~~
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
ktransformers自己修改了一个定制版本的flashinfer,别装官方那个,装git submodule 下的那个customflashinfer |
Beta Was this translation helpful? Give feedback.
总结:
,没准有奇效