如何在tensorflow的LSTM单元中，增加BN层？
0
[已关闭问题]
关闭于 2019-10-24 17:43
我想定义一个多层LSTM_cell，需要对每层输出增加一个BN层和激活函数，代码如下：
def get_lstm_cell(rnn_size,keep_prob):
lstm_cell = tf.contrib.rnn.LSTMCell(rnn_size,
initializer=tf.truncated_normal_initializer(stddev=0.1,seed=2))
lstm_cell = tf.layers.batch_normalization(lstm_cell,training=True) #这种写法是错的，要怎么改？
lstm_cell = tf.relu(lstm_cell)
return lstm_cell
实际运行会报错，提示：('Input has undefined rank:', TensorShape(None))
该怎么解决呢
