卷积神经网路每次循环结果都不一样，怎样使每次是同一个结果

Question

权重初始化时已经设置了seed为固定值，偏置也定义的为常数
weights=tf.Variable(tf.truncated_normal([3, 1, 1, 64],stddev= 0.1,seed=5))*0.01
bias=f.Variable(tf.constant(0.01,shape=[64]))

Answer 1

https://blog.csdn.net/loadqian/article/details/80596437