question about cal |Δndcg|

[BlackcOVER]

I don't understand why different query in the same batch use the same set of indexes?
model.py line 419
index = tf.reshape(tf.range(1., tf.cast(self.batch_size, dtype=tf.float32) + 1), tf.shape(self.label))

[totoyang]

I also think this is a mistake. index should be sorted according to the score generated by each iteration.