cost:nan about abcnn HOT 4 OPEN

I also encountered this problem when used ABCNN-1 and ABCNN3.
But the problem seems to be different from the one in the issue. The nan occured at the beginning of my training.
I finally find out the problem may be caused by tf.sqrt() when calculating the attention matrix. Specifically, it may generate some values such as -3.0*10^-6 which should be zero.

from abcnn.

我也用的自己的数据集，欧几里得距离换成了L1距离后，就没有nan值产生了。

from abcnn.

The same. The nan occured at the beginning of my training when used ABCNN-1 and ABCNN3.
Use the Manhattan distance can solve this problem.
However, in this case, how can we use the Euclidean distance?

from abcnn.

Adding 1e-6 into tf.sqrt() can run well.

from abcnn.