tensorflow gather_nd实现选取阅读理解任务的start end比如

tensorflow gather_nd实现选取阅读理解任务的start end比如

import tensorflow as tfbatch_size = 3seq_len = 5sequence_tensor = tf.constant([[[1,1],[2,2],[3,3],[4,4],[5,5]],[[6,6],[7,7],[8,8],[9,9],[10,10]],[[11,11],[12,12],[13,13],[14,14],[15,15]]])batch_index = tf.cast(tf.range(0, batch_size), tf.int32)sequence_index = tf.cast(tf.ones([batch_size]),tf.int32)index = tf.stack([batch_index, sequence_index], -1)result = tf.gather_nd(params=sequence_tensor,indices=index)sess = tf.Session()print(sess.run(sequence_tensor))print(sess.run(index))print(sess.run(result))

不误导，想找标准的阅读理解任务代码还是 https://github.com/google-research/bert/blob/master/run_squad.py

版权声明：本文内容由网络用户投稿，版权归原作者所有，本站不拥有其著作权，亦不承担相应法律责任。如果您发现本站中有涉嫌抄袭或描述失实的内容，请联系我们jiasou666@gmail.com 处理，核实后本网站将在24小时内删除侵权内容。