简体   繁体   English

[tensorflow]安装tensorflow-gpu,导入错误为分段错误(核心转储)

[英][tensorflow]install tensorflow-gpu with import error as Segmentation fault (core dumped)

Use the command ' pip install --ingore-installed -- upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.3.0-cp36-cp36m-linux_x86_64.whl ' to install tensorflow-gpu for Python3.6 as official instruction. 使用命令' pip install --ingore-installed -- upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.3.0-cp36-cp36m-linux_x86_64.whl '为安装tensorflow-gpu Python3.6作为官方说明。

https://www.tensorflow.org/install/install_linux

And it seems like installed successfully as shown 好像安装成功,如图所示

安装成功

but when test in python with " import tensorflow ", it comes with error as "Segmentation fault (core dumped)" . 但是在python中使用“ import tensorflow ”进行测试时,它带有错误,如"Segmentation fault (core dumped)" 分段故障

When using the similar to install tensorflow for cpu, there is no problem. 当使用类似的方法为CPU安装tensorflow时,没有问题。

My system is Ubuntu 16.04.2 LTS and my Anaconda env "tf-gpu" is new built, without other python lib to disturb. 我的系统是Ubuntu 16.04.2 LTS,而我的Anaconda env“ tf-gpu”是新建的,没有其他python lib可以打扰。

I didn't find similar problem and solution as import to cause dump. 我没有找到与导致转储的导入类似的问题和解决方案。 How to deal with it? 怎么处理呢?

It's probably because you are using incompatible CUDA and/or cuDNN. 可能是因为您使用了不兼容的CUDA和/或cuDNN。 Follow this to see which version to use. 按照操作查看要使用的版本。

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM