时间戳

首页 > 解决方案 > 安装 Tensorflow-gpu 后出现错误 Failed to load the native TensorFlow runtime

tensorflow - 安装 Tensorflow-gpu 后出现错误 Failed to load the native TensorFlow runtime

问题描述

我已经安装了 redhat 6.6、cuda 9.0、cudnn 7.3.1 和 tesnorflow-gpu。但安装后出现以下错误。我的显卡是 nvidia titan v Traceback（最近一次通话最后一次）：

  File "<ipython-input-1-88d96843a926>", line 1, in <module>
    import keras

  File "/root/anaconda3/lib/python3.7/site-packages/keras/__init__.py", line 3, in <module>
    from . import utils

  File "/root/anaconda3/lib/python3.7/site-packages/keras/utils/__init__.py", line 6, in <module>
    from . import conv_utils

  File "/root/anaconda3/lib/python3.7/site-packages/keras/utils/conv_utils.py", line 9, in <module>
    from .. import backend as K

  File "/root/anaconda3/lib/python3.7/site-packages/keras/backend/__init__.py", line 89, in <module>
    from .tensorflow_backend import *

  File "/root/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py", line 5, in <module>
    import tensorflow as tf

  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/__init__.py", line 24, in <module>
    from tensorflow.python import pywrap_tensorflow  # pylint: disable=unused-import

  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/__init__.py", line 49, in <module>
    from tensorflow.python import pywrap_tensorflow

  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 74, in <module>
    raise ImportError(msg)

ImportError: Traceback (most recent call last):
  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 58, in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 24, in swig_import_helper
    _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
  File "/root/anaconda3/lib/python3.7/imp.py", line 242, in load_module
    return load_dynamic(name, filename, file)
  File "/root/anaconda3/lib/python3.7/imp.py", line 342, in load_dynamic
    return _load(spec)
ImportError: libcuda.so.1: cannot open shared object file: No such file or directory

标签： tensorflow

解决方案

您是否测试过您的 cuda 和 cudnn 安装：

英伟达驱动：

$ nvidia-smi

这应该会向您显示您的 GPU 指标的概览

测试cuda：

首先检查是否：

$ nvcc -V

显示您的 cuda 工具包的正确版本

然后您可以使用以下过程对其进行测试：首先：

 $ cd ~/NVIDIA_CUDA-9.0_Samples
 $ make

接着：

$ cd ~/NVIDIA_CUDA-10.0_Samples/bin/x86_64/linux/release
$./deviceQuery

如果你最后有一个“结果：通过”，你就很好！

要测试 cudnn：

$ cp -r /usr/src/cudnn_samples_v7/ $HOME
$ cd $HOME/cudnn_samples_v7/mnistCUDNN
$ make clean && make
$ ./mnistCUDNN

结果应该是：“测试通过！”

张量流：

如果 cuda 和 cudnn 工作正常，我建议您在 conda 环境中安装 tensorflow，使用：

conda create --name tf_gpu tensorflow-gpu

对我来说（在遇到很多问题之后），它运行得很好。

要测试它：

from tensorflow.python.client import device_lib
device_lib.list_local_devices()

主要来源：为 GPU 安装 tensorflow GPU 变得容易

推荐阅读