tensorflow打印内部张量

1.training_decoder_output保存了dynamic_decoder过程的结果，其形式为tuple（rnn_output,sample_id）
traing_logits获取了training_decoder_output中的rnn_output
k1获取shape,k2获取具体traing_logitsֵ

training_decoder_output, _, _ = tf.contrib.seq2seq.dynamic_decode(training_decoder,                                                                        impute_finished=True,                                                                        maximum_iterations=max_target_sequence_length)

training_logits = tf.identity(training_decoder_output.rnn_output, 'logits')

k1=tf.shape(training_logits) 

k2=training_logits

2.在Session中，feed数据，并打印training_logits的值

 m1,m2=sess.run([k1,k2],{input_data: sources_batch,                 targets: targets_batch,                  lr: learning_rate,                  target_sequence_length: targets_lengths,                  source_sequence_length: sources_lengths}

  print('logits',m1)

文章来源: tensorflow打印内部张量