Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, 4

处理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. [Op:Conv2D]报错
tensorflow2.0

import tensorflow as tf
import keras
rand_input = tf.random.normal([20, 50, 100, 16]) #输入参数有所区别[N(batch_size),h,w,in_chanles]
m = tf.keras.layers.Conv2D(33, 3, strides=2)
output = m(rand_input)
print(output.shape)

报错：
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. [Op:Conv2D]

找了很多方法，有人说是cudnn什么的版本不匹配，有人说是显存爆炸（我觉得我的原因应该不是这个，毕竟这几段代码不需要多少内存吧，我也不太懂其实，实在是太白了）。只有最后一个方法有用！

import os
os.environ['CUDA_VISIBLE_DEVICES'] = '/gpu:0'

虽然还是不知道原因和原理，但是最近真的是太忙了，之后有时间的话就搞搞清楚吧！