tensorflow-gpu1.14 + Win10 + CUDA10.0 + CUDNN7.5.0 + Python3.6 + VS2015安装

最近学习深度学习，在配置环境中的过程中遇到很多问题，在这进行总结，希望对大家有帮助。

一、整个软件安装配置过程，很多博客写的很详细，附上链接；

1. https://blog.csdn.net/qq_36026791/article/details/88793488
2. https://blog.csdn.net/huanyingzhizai/article/details/89298964

二、安装完成后测试如下；

1. 整个tensorflow测试过程

   (base) C:\Users\Wang>activate tensorflow--gpu//激活tensorflow环境
   
   (tensorflow--gpu) C:\Users\Wang>python
   Python 3.6.10 |Anaconda, Inc.| (default, Jan  7 2020, 15:18:16) [MSC v.1916 64 bit (AMD64)] on win32
   Type "help", "copyright", "credits" or "license" for more information.
   >>> from tensorflow.python.client import device-lib
     File "<stdin>", line 1
       from tensorflow.python.client import device-lib
                                                  ^
   SyntaxError: invalid syntax
   >>> from tensorflow.python.client import device_lib
   >>> print(device_lib.list_local_devices())
   2020-03-08 09:44:53.011007: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
   2020-03-08 09:44:53.892132: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1433] Found device 0 with properties:
   name: GeForce MX150 major: 6 minor: 1 memoryClockRate(GHz): 1.0375
   pciBusID: 0000:01:00.0
   totalMemory: 2.00GiB freeMemory: 1.62GiB
   2020-03-08 09:44:53.914731: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1512] Adding visible gpu devices: 0
   2020-03-08 09:44:55.706053: I tensorflow/core/common_runtime/gpu/gpu_device.cc:984] Device interconnect StreamExecutor with strength 1 edge matrix:
   2020-03-08 09:44:55.722874: I tensorflow/core/common_runtime/gpu/gpu_device.cc:990]      0
   2020-03-08 09:44:55.731764: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1003] 0:   N
   2020-03-08 09:44:55.746369: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/device:GPU:0 with 1364 MB memory) -> physical GPU (device: 0, name: GeForce MX150, pci bus id: 0000:01:00.0, compute capability: 6.1)
   [name: "/device:CPU:0"
   device_type: "CPU"       //抓到CPU
   memory_limit: 268435456
   locality {
   }
   incarnation: 17524159819158732381
   , name: "/device:GPU:0"
   device_type: "GPU"     //抓到GPU
   memory_limit: 1431112499
   locality {
     bus_id: 1
     links {
     }
   }
   incarnation: 5191225488240913321
   physical_device_desc: "device: 0, name: GeForce MX150, pci bus id: 0000:01:00.0, compute capability: 6.1"
   ]
   >>> import tensorflow as tf
   >>> gjm = tf.constant("Hello World ! I love TensorFlow ! ")
   >>> sess = tf.Session()
   2020-03-08 09:46:10.621388: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1512] Adding visible gpu devices: 0
   2020-03-08 09:46:11.108336: I tensorflow/core/common_runtime/gpu/gpu_device.cc:984] Device interconnect StreamExecutor with strength 1 edge matrix:
   2020-03-08 09:46:11.124001: I tensorflow/core/common_runtime/gpu/gpu_device.cc:990]      0
   2020-03-08 09:46:11.133490: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1003] 0:   N
   2020-03-08 09:46:11.264295: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 1364 MB memory) -> physical GPU (device: 0, name: GeForce MX150, pci bus id: 0000:01:00.0, compute capability: 6.1)
   >>> print (sess.run(gjm))
   2020-03-08 09:46:44.969417: E tensorflow/core/grappler/clusters/utils.cc:83] Failed to get device properties, error code: 30
   b'Hello World ! I love TensorFlow ! '    //输出
   >>> sess.close()
   >>>

    

2. 查看cuda是否安装成功
3. 其他过程参见链接博客

三、遇到的几个问题；

1. 首先是在安装过程中，python、tensorflow、Cuda版本的对应，这里我的版本如下：python3.6（在安装Anaconda3时候自带为python3.7，这里在Anaconda Prompt里执行降版本，降为3.6即可）、tensorflow1.14rc0、Cuda 10.0。如果版本有问题，不兼容，后面测试会一直报错。
2. Cuda 10.0安装完成后，一定要配置环境变量，具体如下：检查一下环境变量 我的电脑—> 属性—> 高级系统设置—> 环境变量—>系统变量      

   
   确认环境变量，CUDA_PATH和CUDA_PATH_V10已经存在

   依此添加下列三个路径到Path里
   C:\NVIDIA\CUDAv10.1\bin
   C:\NVIDIA\CUDAv10.1\include
   C:\NVIDIA\CUDAv10.1\lib\x64
   否则在Windows CMD命令窗口执行 nvcc -V 的时候，会发生错误，查不到Cuda版本信息，并且在后续测试Tensorflow的时候，会发生如下错误找不到指定模块

3. 安装tensorflow，如果直接使用镜像，下载速度非常慢，而且容易出现错误。这里推荐使用清华镜像pip install -i https://pypi.tuna.tsinghua.edu.cn/simple tensorflow-gpu==1.14.0rc0
4. 还有一点，在安装完tensorflow后，一般会同时安装numpy，但是自带的numpy版本过高，会出现以下错误

这里对numpy进行降版本操作就行。

就是先卸载原来的numpy     pip uninstall numpy 然后用清华镜像重新安装pip install -i https://pypi.tuna.tainghua.edu.cn/simple numpy==1.16.0

以上是我遇到的一些问题及坑的解决方法，欢迎大家参考，有问题留言交流。