Opened this issue 6 years ago · 1 comments
既然喂给网络一个batch的数据,计算的不应该是一个batch的损失吗?为什么后面计算平均损失的时候需要再乘以batch_size?
@yyHaker I agree with you,maybe he intended to use "tf.reduce_mean" instead of "tf.reduce_sum"