Question

我尝试在Ubuntu 17.10上安装tensorflow，并且我正在完成本指南的组合：https://medium.com/@adakaminkure/if-youre-trying-to-install-tensorflow-on-ubuntu-17-10-703b971d98d

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

我遇到了一些问题。

我坚持使用mnist编译

$ cp -r / usr / src / cudnn_samples_v7 / $ HOME

$ cd $ HOME / cudnn_samples_v7 / mnistCUDNN

$ make

$ ./mnistCUDNN

我收到以下错误：

cudnnGetVersion() : 7102 , CUDNN_VERSION from cudnn.h : 7102 (7.1.2)
Cuda failurer version : GCC 6.4.0
Error: unknown error
error_util.h:93
Aborting...

如果我移过那个并尝试在virtualenv上安装tensorflow：

$ sudo apt-get install python3-pip python3-dev python-virtualenv

$ virtualenv --system-site-packages -p python3~ / tensorflow

$ source~ / tensorflow / bin / activate

$ easy_install -U pip

$ pip3 install --upgrade tensorflow -gpu

我在最后的命令输出中注意到以下内容：

protobuf 3.5.2.post1 requires six>=1.9, which is not installed.
grpcio 1.10.1 requires six>=1.5.2, which is not installed.
absl-py 0.1.13 requires six, which is not installed.
html5lib 0.9999999 requires six, which is not installed.
bleach 1.5.0 requires six, which is not installed.
tensorboard 1.7.0 requires six>=1.10.0, which is not installed.
tensorflow-gpu 1.7.0 requires six>=1.10.0, which is not installed.

然后我试着跑：

$ import tensorflow as tf

$ hello = tf.constant（＆＃34;你好，来自Tensorflow的Ada＆＃34;）

$ session = tf.Session（）

我收到以下错误：

2018-04-08 16:33:31.297906: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2018-04-08 16:33:31.346416: E tensorflow/stream_executor/cuda/cuda_driver.cc:406] failed call to cuInit: CUDA_ERROR_UNKNOWN
2018-04-08 16:33:31.346468: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:145] kernel driver does not appear to be running on this host (spacecowboy): /proc/driver/nvidia/version does not exist

知道发生了什么事吗？

在Ubuntu 17.10上安装Tensorflow时遇到问题

0 个答案: