安装张量流 1.5 后出错（linux mint 18.3）

Question

刚刚完成张量流（gpu 启用）1.5 的安装，在 Clevo 笔记本电脑上，在 linux mint 18.3 和 nvidia 1060 gpu 下运行测试程序 python premade_estimator.py 我收到以下错误代码：

    Traceback (most recent call last):
  File "premade_estimator.py", line 20, in <module>
    import tensorflow as tf
  File "/home/empereur/tensorflow/local/lib/python2.7/site-packages/tensorflow/__init__.py", line 24, in <module>
    from tensorflow.python import *
  File "/home/empereur/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/__init__.py", line 49, in <module>
    from tensorflow.python import pywrap_tensorflow
  File "/home/empereur/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 74, in <module>
    raise ImportError(msg)
ImportError: Traceback (most recent call last):
  File "/home/empereur/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 58, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "/home/empereur/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "/home/empereur/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 24, in swig_import_helper
    _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
ImportError: libcublas.so.9.0: cannot open shared object file: No such file or directory


Failed to load the native TensorFlow runtime.

See https://www.tensorflow.org/install/install_sources#common_installation_problems

for some common reasons and solutions.  Include the entire stack trace
above this error message when asking for help.

我已经安装了 cuda 驱动程序 9.0，如图所示：

nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Sep__1_21:08:03_CDT_2017
Cuda compilation tools, release 9.0, V9.0.176

我查看了论坛，大多数问题要么是两个版本之间的冲突（不是我的情况），要么是版本 9.0 的路径没有完成（我想我已经完成）

我对此有点新手，所以非常感谢帮助

如果缺少任何信息，请询问，我会提供信息（但拜托，我是新手，所以让我知道如何使用命令行获取信息）

这里是usr/local/

的内容

empereur@empereur-Sky-X4C ~ $ ls /usr/local/ bin cuda
cuda-9.1 games lib sbin src Brother cuda-9.0 etc
include man share

下面是usr/local/cuda/lib64的内容libaccinj64.so

empereur@empereur-Sky-X4C ~ $ ls /usr/local/cuda/lib64 libaccinj64.so libcusparse.so.9.1.85 libnppist.so.9.1.85 libaccinj64.so.9.1
libcusparse_static.a libnppist_static.a libaccinj64.so.9.1.85 libnppc.so libnppisu.so libcublas_device.a
libnppc.so.9.1 libnppisu.so.9.1 libcublas.so
libnppc.so.9.1.85 libnppisu.so.9.1.85 libcublas.so.9.1
libnppc_static.a libnppisu_static.a libcublas.so.9.1.128
libnppial.so libnppitc.so libcublas.so.9.1.85
libnppial.so.9.1 libnppitc.so.9.1 libcublas_static.a
libnppial.so.9.1.85 libnppitc.so.9.1.85 libcudadevrt.a
libnppial_static.a libnppitc_static.a libcudart.so
libnppicc.so libnpps.so libcudart.so.9.1
libnppicc.so.9.1 libnpps.so.9.1 libcudart.so.9.1.85
libnppicc.so.9.1.85 libnpps.so.9.1.85 libcudart_static.a
libnppicc_static.a libnpps_static.a libcufft.so
libnppicom.so libnvblas.so libcufft.so.9.1
libnppicom.so.9.1 libnvblas.so.9.1 libcufft.so.9.1.85
libnppicom.so.9.1.85 libnvblas.so.9.1.128 libcufft_static.a
libnppicom_static.a libnvblas.so.9.1.85 libcufftw.so
libnppidei.so libnvgraph.so libcufftw.so.9.1
libnppidei.so.9.1 libnvgraph.so.9.1 libcufftw.so.9.1.85
libnppidei.so.9.1.85 libnvgraph.so.9.1.85 libcufftw_static.a
libnppidei_static.a libnvgraph_static.a libcuinj64.so
libnppif.so libnvrtc-builtins.so libcuinj64.so.9.1
libnppif.so.9.1 libnvrtc-builtins.so.9.1 libcuinj64.so.9.1.85
libnppif.so.9.1.85 libnvrtc-builtins.so.9.1.85 libculibos.a
libnppif_static.a libnvrtc.so libcurand.so libnppig.so libnvrtc.so.9.1 libcurand.so.9.1 libnppig.so.9.1
libnvrtc.so.9.1.85 libcurand.so.9.1.85 libnppig.so.9.1.85
libnvToolsExt.so libcurand_static.a libnppig_static.a
libnvToolsExt.so.1 libcusolver.so libnppim.so
libnvToolsExt.so.1.0.0 libcusolver.so.9.1 libnppim.so.9.1
libOpenCL.so libcusolver.so.9.1.85 libnppim.so.9.1.85
libOpenCL.so.1 libcusolver_static.a libnppim_static.a
libOpenCL.so.1.0 libcusparse.so libnppist.so
libOpenCL.so.1.0.0 libcusparse.so.9.1 libnppist.so.9.1 stubs

Answer 1

感谢大家的帮助，所以我不小心安装了cuda 9.1（需要9.0）。我这样做并没有特别强制使用 cuda-toolkits 上的版本。通过清除 cuda 9.1 和 nvidia-cuda-toolkit 并重新安装 9.0 版本（不正确地）解决了我的问题。按照 laylog

的建议，可能可以用更干净的方式解决它

But instead of purging 9.1, you can point the cuda symlink to point to CUDA-9.0 or update the LD_LIBRARY_PATH in your bashrc to point to CUDA-9.0

再次感谢大家！

安装张量流 1.5 后出错（linux mint 18.3）

Error after install of tensor flow 1.5 (linux mint 18.3)

python

linux-mint

tensorflow