Plug-and-Play version of Attention implemented in Tensorflow (1.2.1).
Install Tensorflow for your system.
You can use this file with your code in the following manner :
from attention import Attention
...
output_vectors = lstm()
atta = Attention(output_vectors)
att_vec = atta.applyAttention()
output = tf.nn.softmax(tf.matmul(Wh, att_vec) + b)
...Haven't checked, will be running it on something soon, hopefully.