1、Colab: (0) UNIMPLEMENTED: DNN library is not found
2、Unknown: Failed to get convolution algorithm. This is probably because cuDNN:
添加下面两行
import os
os.environ[‘CUDA_VISIBLE_DEVICES’] = ‘/gpu:0’
方法一:可能是Tensorflow-gpu版本太高,我报错时为1.12,可以将其降为1.9.0
pip install tensorflow-gpu1.9.0或者conda install tensorflow-gpu1.9.0
Note: 我不知道是不是同时需要将cpu版本的tensorflow也降下来,如果你还有问题,则继续执行
pip install tensorflow1.9.0或者conda install tensorflow1.9.0