解决This is probably because cuDNN failed to initial

报错内容

tensorflow.python.framework.errors_impl.UnknownError:  Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.

网上文章都是说cudnn版本不匹配，但是查了之后发现不是这个问题。

https://www.tensorflow.org/in...

然后看到了一篇博客说到可能是因为显存不足，导致的。

限制GPU的使用率

方法一

config = tf.ConfigProto()
config.gpu_options.per_process_gpu_memory_fraction = 0.85  #占用85%显存
session = tf.Session(config=config)

方法二

gpu_options=tf.GPUOptions(per_process_gpu_memory_fraction=0.85)
config=tf.ConfigProto(gpu_options=gpu_options)
session = tf.Session(config=config)

参考文章

矩池云中Tensorflow指定GPU及GPU显存设置

“Failed to get convolution algorithm. This is probably because cuDNN failed to initialize”错误的解决办法