WARNING:tensorflow:The name tf.GPUOptions is deprecated. Please use tf.compat.v1.GPUOption.解决方案

在ubuntu系统上使用tensorflow-gpu1.14.0版本运行深度学习模型每次弹出的warning:

WARNING:tensorflow:From execute.py:233: The name tf.set_random_seed is deprecated. Please use tf.compat.v1.set_random_seed instead.

From execute.py:233: The name tf.set_random_seed is deprecated. Please use tf.compat.v1.set_random_seed instead.

WARNING:tensorflow:From execute.py:234: The name tf.GPUOptions is deprecated. Please use tf.compat.v1.GPUOptions instead.

From execute.py:234: The name tf.GPUOptions is deprecated. Please use tf.compat.v1.GPUOptions instead.

WARNING:tensorflow:From execute.py:235: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.

From execute.py:235: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.

WARNING:tensorflow:From execute.py:236: The name tf.Session is deprecated. Please use tf.compat.v1.Session instead.

From execute.py:236: The name tf.Session is deprecated. Please use tf.compat.v1.Session instead.

解决方案:

系统:ubuntu, anaconda环境

卸载当前tensorflow-gpu版本

pip uninstall tensorflow-gpu

安装tensorflow-gpu 1.5.0版本

pip install tensorflow-gpu==1.50

可能会出现以下错误:

这时换成国内源就可以了

pip install tensorflow-gpu==1.5.0 -i https://pypi.douban.com/simple

更换tensorflow版本可能出现的问题:

ImportError: libcublas.so.9.0: cannot open shared object file: No such file or directory
这就是tensorflow版本与cuda版本不兼容。我的是cuda10 。可以看出cuda 10与tensorflow 1.5.0不兼容, 上面错误提示 libcublas.so.9.0的意思就是需要更换成cuda 9.0才能运行。

tensorflow版本与cuda版本匹配信息:来源于csdn博客

解决方法:

一是更换cuda版本,二是换tensorflow版本

猜你喜欢

转载自blog.csdn.net/Thanours/article/details/102865283