mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2026-11-05 01:07:28 +00:00
[GH-ISSUE #1088] v0.2.4 post1 编译失败 #766
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
kvcache-ai/ktransformers#766
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 @TTThanos on GitHub (Apr 8, 2025).
Original GitHub issue: https://github.com/kvcache-ai/ktransformers/issues/1088
今天pull了最新版的KT,按照https://github.com/kvcache-ai/ktransformers/blob/main/doc/en/balance-serve.md 中的安装步骤,当运行至

USE_BALANCE_SERVE=1 bash ./install.sh时,报以下错误,请问该怎么解决CUDA 12.4
Driver 550
Python 3.10
@ChachAloha commented on GitHub (Apr 8, 2025):
我环境之前2.3版本也都可以,24开始就不行的重新开了conda,补了依赖也一样,他们说只有官方的容器可以正常编译
@magikRUKKOLA commented on GitHub (Apr 8, 2025):
Please upgrade your drivers, cuda, python, gcc, g++, gcc and g++ dev libraries outside of your envirtonment too, and everything else! Welcome to the World of Python, the world of ultimate convenience and portability!
https://github.com/kvcache-ai/ktransformers/issues/1051#issuecomment-2785168929
@magikRUKKOLA commented on GitHub (Apr 8, 2025):
I am sorry, but its simply not true. You can compile its all on your own provided your kept the versions of the software inside and outside of your python environment. I have put the link above -- its totally doable at least on Debian 12. Alternatively, you can listen to the professionals in the field. For example ... https://www.youtube.com/watch?v=YnL9vAFphmE
@TTThanos commented on GitHub (Apr 9, 2025):
我用了KT官方的容器build KT也还是不行,容器拉取命令
docker pull approachingai/ktransformers:v0.2.4-AVX512@shinchou commented on GitHub (Apr 9, 2025):
sudo apt update && sudo apt install gcc g++ gcc-12 g++-12 build-essential cmake
update-alternatives --install /usr/bin/cc cc /usr/bin/gcc-12 100
update-alternatives --install /usr/bin/c++ c++ /usr/bin/g++-12 100
update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-12 100
update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-12 100
@Zhiwei-Dai commented on GitHub (Apr 10, 2025):
也遇到了这个问题,0.2.3正常,升级到0.2.4报出一堆异常,也定位不到原因。我做了两点尝试,通过了,供参考:
1、cmake可使用pip安装升级到最新4.0
2、看了B站的视频之后,参考脚本执行正常:env USE_BALANCE_SERVE=1 USE_NUMA=1 PYTHONPATH="$(which python)" PATH="$(dirname $(which python)):$PATH" bash ./install.sh。(参考官方的就是不行:USE_BALANCE_SERVE=1 USE_NUMA=1 bash ./install.sh)
@TTThanos commented on GitHub (Apr 11, 2025):
我目前解决了问题,按照https://github.com/kvcache-ai/ktransformers/blob/main/doc/en/balance-serve.md 安装指示。
pull了镜像,创建python 3.11 conda环境,安装依赖,重要的是 重新安装了cpufeature到0.2.2版本, 安装flash attention,conda内安装cmake,之后就可以用
USE_BALANCE_SERVE=1 bash ./install.shbuild 0.2.4 KT了。机器四卡A10
Ubuntu 22.04
驱动版本570
CUDA 12.8
@wishfay commented on GitHub (Apr 23, 2025):
感谢!这样确实可以跑起来了。
需要手动装一下cmake,其实应该算是那个requirement.txt文件少写了一个cmake,如果加上应该也是可以的。