有很多方法可以得出非限定结果

There are a lot of ways you can end up with non-finite results.

但是，如果学习率过高，则优化器（尤其是诸如梯度下降之类的简单优化器）可能会发散。

But optimizers, especially simple ones like gradient descent, can diverge if the learning rate is 'too high'.

您是否尝试过简单地将学习率除以10/100/1000？或通过 pixels * batch_size 进行标准化以获得每个像素的平均误差？

Have you tried simply dividing your learning rate by 10/100/1000? Or normalizing by pixels*batch_size to get the average error per pixel?

还是更多高级优化程序？例如，带有默认选项的 tf.train.AdamOptimizer（）。

Or one of the more advanced optimizers? For example tf.train.AdamOptimizer() with default options.