ubuntu anaconda cuda8.0 cudnn5.1.0 py3.6 tensorflow-gpu

雨点打透心脏的1/2处 2022-02-26 14:28 138阅读 0赞

### 安装anaconda ###

参考 [https://blog.csdn.net/weixin\_37485708/article/details/86416208][https_blog.csdn.net_weixin_37485708_article_details_86416208]

### 安装cuda8.0 和cudnn5.1 ###

参考 [https://blog.csdn.net/weixin\_37485708/article/details/82794326][https_blog.csdn.net_weixin_37485708_article_details_82794326]

### 安装tensorflow==1.2.1 ###

参考 这个 [https://blog.csdn.net/shiheyingzhe/article/details/80863422][https_blog.csdn.net_shiheyingzhe_article_details_80863422] 借图片一用。所以我们要安装的tensorflow是1.2.1版本。

![70][]

创建虚拟环境  conda create -n tf-gpu python=3.6

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70][]

激活虚拟环境 source activate tf-gpu

conda install tensorflow-gpu==1.2.1  
然后就慢慢等着咯

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 1][]

测试是否安装成功

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 2][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 3][]

import tensorflow  
>>> import numpy  
>>> import tensorflow as tf  
>>> a=tf.constant(\[1.0,2.0,3.0,4.0,5.0,6.0\],shape=\[2,3\],name='a')  
>>> b=tf.constant(\[1.0,2.0,3.0,4.0,5.0,6.0\],shape=\[3,2\],name='b')  
>>> c=tf.matmul(a,b)  
>>> sess=tf.Session(config=tf.ConfigProto(log\_device\_placement=True))  
>>> print(sess.run(c))  
这个方法还可以查看是cpu版本的tensorflow还是gpu版本的  参考于 [https://blog.csdn.net/weixin\_37251044/article/details/79790270][https_blog.csdn.net_weixin_37251044_article_details_79790270]

### 安装keras ###

参考 [https://blog.csdn.net/Super\_jm\_/article/details/78397074\#commentBox][https_blog.csdn.net_Super_jm_article_details_78397074_commentBox] （这篇博客超级全！！！）

pip install keras

### pycharm ###

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 4][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 5][]

### keras\\opencv ###

keras是直接通过anaconda的图形化界面安装的，在终端中输入 anaconda-navigator 可以打开图像管理界面，选到tf-gpu虚拟环境下

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 6][]

选择all，可以直接通过图形化界面安装。当然也可以直接pip install

但是opencv我试了下，在图形化界面中安装失败了。

于是 [https://blog.csdn.net/weixin\_37485708/article/details/86416208][https_blog.csdn.net_weixin_37485708_article_details_86416208] 我之前写的博文，先下载好了配套的opencv包，然后安装。

参考：

[https://blog.csdn.net/qq\_23888879/article/details/80630539][https_blog.csdn.net_qq_23888879_article_details_80630539] （说了安装错怎么办）

[https://blog.csdn.net/lukaslong/article/details/81092032][https_blog.csdn.net_lukaslong_article_details_81092032]  超级简单方便！

[https_blog.csdn.net_weixin_37485708_article_details_86416208]: https://blog.csdn.net/weixin_37485708/article/details/86416208
[https_blog.csdn.net_weixin_37485708_article_details_82794326]: https://blog.csdn.net/weixin_37485708/article/details/82794326
[https_blog.csdn.net_shiheyingzhe_article_details_80863422]: https://blog.csdn.net/shiheyingzhe/article/details/80863422
[70]: /images/20220226/9f431a91fa734a288fec41f75f54ba66.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70]: /images/20220226/342e9a8bf9ed42349f216bd0f8b9b964.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 1]: /images/20220226/3747769e86a3416daa558f84eba07fbf.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 2]: /images/20220226/20b8d90e8a35405986b6e1178a8017cd.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 3]: /images/20220226/3954cf00c114431d8295566e67820fce.png
[https_blog.csdn.net_weixin_37251044_article_details_79790270]: https://blog.csdn.net/weixin_37251044/article/details/79790270
[https_blog.csdn.net_Super_jm_article_details_78397074_commentBox]: https://blog.csdn.net/Super_jm_/article/details/78397074#commentBox
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 4]: /images/20220226/376dcf369f924c338827a3819e230ab4.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 5]: /images/20220226/e9a6288ec5bb4b0e8aac649771dfc9a2.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl8zNzQ4NTcwOA_size_16_color_FFFFFF_t_70 6]: /images/20220226/e238be499a0e47c9867ed6c94e3ac251.png
[https_blog.csdn.net_qq_23888879_article_details_80630539]: https://blog.csdn.net/qq_23888879/article/details/80630539
[https_blog.csdn.net_lukaslong_article_details_81092032]: https://blog.csdn.net/lukaslong/article/details/81092032