one question about top_antecedent_cluster_ids

kentonl / e2e-coref

End-to-end Neural Coreference Resolution

Apache License 2.0

518 stars 174 forks source link

one question about top_antecedent_cluster_ids #74

Open MENGHAH opened 4 years ago

MENGHAH commented 4 years ago

in coref_model.py top_antecedent_cluster_ids = tf.gather(top_span_cluster_ids, top_antecedents) top_antecedent_cluster_ids += tf.to_int32(tf.log(tf.to_float(top_antecedents_mask)))

The first step have got the top_antecedent_cluster_ids,why does there still run the second step?