Windows10搭建tensorflow-gpu2.1环境

谁践踏了优雅 2023-07-11 11:07 157阅读 0赞

### 一、安装Anaconda3 ###

> Anaconda是一个开源的Python包管理器，其包含了conda、Python等180多个科学包及其依赖项，非常实用。

##### 1. 下载 #####

下载可以直接到Anaconda3官网：[https://www.anaconda.com/distribution][https_www.anaconda.com_distribution]  
附国内清华镜像网：[https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/?C=M&O=D][https_mirrors.tuna.tsinghua.edu.cn_anaconda_archive_C_M_O_D]  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70]

##### 2. 安装 #####

没什么好说的，要注意的一点就是要勾选这一项，自动配置环境变量  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 1]

##### 3. 验证 #####

装好之后打开cmd，输入 conda --version  
显示如下即为安装成功  
![在这里插入图片描述][20200305094909216.png]

### 二、安装CUDA ###

##### 1.下载CUDA10.1（官网）： #####

[https://developer.nvidia.com/cuda-10.1-download-archive-update2?target\_os=Windows&target\_arch=x86\_64&target\_version=10&target\_type=exelocal][https_developer.nvidia.com_cuda-10.1-download-archive-update2_target_os_Windows_target_arch_x86_64_target_version_10_target_type_exelocal]  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 2]

##### 2.安装CUDA10.1 #####

下面两项不装，其它全部默认即可（保证所装版本＞旧版本）  
![在这里插入图片描述][20200305111253245.png]![在这里插入图片描述][20200305111307306.png]  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 3]

##### 3.检查CUDA是否安装成功 #####

*  打开环境变量配置界面，检查是否有这两项  
    ![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 4]
 *  在命令行中输入nvcc --version查看安装的版本
 *  查看C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v10.1\\bin中是否存在nvcc.exe
 *  查看C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v10.1\\extras\\CUPTI\\lib64是否存在cupti64\_101.dll

**符合以上条件即视为CUDA10.1安装成功。**

### 三、安装cudnn ###

##### 1. 下载cudnn（官网注册/登陆后下载）： #####

[https://developer.nvidia.com/rdp/cudnn-download][https_developer.nvidia.com_rdp_cudnn-download]  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 5]

##### 2. 解压压缩包至指定目录 #####

将下载到的压缩包解压到 cuda 的安装路径C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v10.1下。  
**注意解压后的文件夹名为cuda，请将其改成cudnn。**  
![在这里插入图片描述][20200305113226776.png]

##### 3. 配置环境变量 #####

确保path变量中有如下四项，如缺少请手动补全  
![在这里插入图片描述][20200305113729672.png]

### 四、安装tensorflow2.1 ###

##### 1. 创建环境 #####

在命令行中输入

conda create -n tf2.1 python==3.7.6

检查是否创建成功输入

conda-env list

如下图所示  
![在这里插入图片描述][20200305120752584.png]

##### 2. 安装tensorflow2.1 #####

命令行中输入如下进入该环境

conda activate tf2.1

*注意：若提示未初始化shell，则按提示输入conda init后重启命令行即可。*

开始安装tensorflow2.1

pip install tensorflow-gpu==2.1

![在这里插入图片描述][20200305121004339.png]

##### 3. 测试 #####

打开 jupyter notebook / pycharm 输入

import tensorflow as tf

测试一下是否成功即可

[https_www.anaconda.com_distribution]: https://www.anaconda.com/distribution
[https_mirrors.tuna.tsinghua.edu.cn_anaconda_archive_C_M_O_D]: https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/?C=M&O=D
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20200305094557190.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 1]: https://img-blog.csdnimg.cn/20200305094738375.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs,size_16,color_FFFFFF,t_70
[20200305094909216.png]: https://img-blog.csdnimg.cn/20200305094909216.png
[https_developer.nvidia.com_cuda-10.1-download-archive-update2_target_os_Windows_target_arch_x86_64_target_version_10_target_type_exelocal]: https://developer.nvidia.com/cuda-10.1-download-archive-update2?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exelocal
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 2]: https://img-blog.csdnimg.cn/20200305102303578.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs,size_16,color_FFFFFF,t_70
[20200305111253245.png]: https://img-blog.csdnimg.cn/20200305111253245.png
[20200305111307306.png]: https://img-blog.csdnimg.cn/20200305111307306.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 3]: https://img-blog.csdnimg.cn/20200305111434331.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 4]: https://img-blog.csdnimg.cn/20200305111721474.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs,size_16,color_FFFFFF,t_70
[https_developer.nvidia.com_rdp_cudnn-download]: https://developer.nvidia.com/rdp/cudnn-download
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs_size_16_color_FFFFFF_t_70 5]: https://img-blog.csdnimg.cn/2020030511281453.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L1NlbnJlbmRpcGl0eUFs,size_16,color_FFFFFF,t_70
[20200305113226776.png]: https://img-blog.csdnimg.cn/20200305113226776.png
[20200305113729672.png]: https://img-blog.csdnimg.cn/20200305113729672.png
[20200305120752584.png]: https://img-blog.csdnimg.cn/20200305120752584.png
[20200305121004339.png]: https://img-blog.csdnimg.cn/20200305121004339.png