CUDA driver version is insufficient for CUDA runtime version

用 conda 安装完 tensorflow-gpu : （注意：conda 自动帮你装了最新的 cuda10！）

>>> conda install tensorflow-gpu
Solving environment: done

## Package Plan ##

  environment location: D:\ProgramFilesNoSpace\Miniconda3\envs\py36

  added / updated specs:
    - tensorflow-gpu


The following NEW packages will be INSTALLED:

    cudatoolkit:     10.0.130-0                https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    cudnn:           7.3.1-cuda10.0_0          https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow:      1.13.1-gpu_py36h9006a92_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow-base: 1.13.1-gpu_py36h871c8ca_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow-gpu:  1.13.1-h0d30ee6_0         https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main

Proceed ([y]/n)? n  # 这个 n 是后来加的

一跑程序就出现：

Traceback (most recent call last):
  File "D:/workspace/pythonProj/test2.py", line 61, in <module>
    sess = tf.Session()
  File "D:\ProgramFilesNoSpace\Miniconda3\envs\py36\lib\site-packages\tensorflow\python\client\session.py", line 1551, in __init__
    super(Session, self).__init__(target, graph, config=config)
  File "D:\ProgramFilesNoSpace\Miniconda3\envs\py36\lib\site-packages\tensorflow\python\client\session.py", line 676, in __init__
    self._session = tf_session.TF_NewSessionRef(self._graph._c_graph, opts)
tensorflow.python.framework.errors_impl.InternalError: cudaGetDevice() failed. Status: CUDA driver version is insufficient for CUDA runtime version

看看 conda 安装的 cuda 版本

>>> conda list
# packages in environment at D:\ProgramFilesNoSpace\Miniconda3\envs\py36:
#
# Name                    Version                   Build  Channel
...
cudatoolkit               10.0.130                      0    https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
cudnn                     7.3.1                cuda10.0_0    https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main

检查显卡驱动版本：

>>> nvidia-smi
Fri Mar 29 11:54:57 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 399.24                 Driver Version: 399.24                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce MX150      WDDM  | 00000000:01:00.0 Off |                  N/A |
| N/A   35C    P8    N/A /  N/A |     64MiB /  2048MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

版本为 399

再一查 tensorflow 系统要求：
https://www.tensorflow.org/install/gpu#software_requirements

搞什么嘛，只支持到 cuda 9.0

只好把 cuda 给卸了

>>> conda uninstall cudnn
Solving environment: done

## Package Plan ##

  environment location: D:\ProgramFilesNoSpace\Miniconda3\envs\py36

  removed specs:
    - cudnn


The following packages will be REMOVED:

    cudnn:           7.3.1-cuda10.0_0          https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow:      1.13.1-gpu_py36h9006a92_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow-base: 1.13.1-gpu_py36h871c8ca_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow-gpu:  1.13.1-h0d30ee6_0         https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main



>>> conda uninstall cudatoolkit
Solving environment: done

## Package Plan ##

  environment location: D:\ProgramFilesNoSpace\Miniconda3\envs\py36

  removed specs:
    - cudatoolkit


The following packages will be REMOVED:

    cudatoolkit: 10.0.130-0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main

现在重装 tensorflow-gpu，选择版本 1.9，对应 cuda9.0

>conda install tensorflow-gpu==1.9
Collecting package metadata: done
Solving environment: done

## Package Plan ##

  environment location: D:\ProgramFilesNoSpace\Miniconda3\envs\py36

  added / updated specs:
    - tensorflow-gpu==1.9


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    cudatoolkit-9.0            |                1       339.8 MB  https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    cudnn-7.3.1                |        cuda9.0_0       170.9 MB  https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorboard-1.9.0          |   py36he025d50_0         3.3 MB  https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow-1.9.0           |gpu_py36hfdee9c2_1           4 KB  https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow-base-1.9.0      |gpu_py36h6e53903_0        78.7 MB  https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    tensorflow-gpu-1.9.0       |       hf154084_0           3 KB  https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
    ------------------------------------------------------------
                                           Total:       592.7 MB

The following NEW packages will be INSTALLED:

  cudatoolkit        anaconda/pkgs/main/win-64::cudatoolkit-9.0-1
  cudnn              anaconda/pkgs/main/win-64::cudnn-7.3.1-cuda9.0_0
  tensorflow         anaconda/pkgs/main/win-64::tensorflow-1.9.0-gpu_py36hfdee9c2_1
  tensorflow-base    anaconda/pkgs/main/win-64::tensorflow-base-1.9.0-gpu_py36h6e53903_0
  tensorflow-gpu     anaconda/pkgs/main/win-64::tensorflow-gpu-1.9.0-hf154084_0

The following packages will be DOWNGRADED:

  tensorboard                         1.13.1-py36h33f27b4_0 --> 1.9.0-py36he025d50_0

发现 conda 还是不错的， 会自动把刚才没有删干净的 tensorboard 自动降级（downgrade）