tf.nn.dynamic_rnn返回两个张量:outputs
and states
.
The outputs
保存批次中所有序列的所有单元的输出。因此,如果特定序列较短并用零填充,则outputs
最后一个单元格将为零。
The states
保存最后一个单元状态,或者等效地每个序列最后一个非零输出(如果您使用BasicRNNCell
).
这是一个例子:
import numpy as np
import tensorflow as tf
n_steps = 2
n_inputs = 3
n_neurons = 5
X = tf.placeholder(dtype=tf.float32, shape=[None, n_steps, n_inputs])
seq_length = tf.placeholder(tf.int32, [None])
basic_cell = tf.nn.rnn_cell.BasicRNNCell(num_units=n_neurons)
outputs, states = tf.nn.dynamic_rnn(basic_cell, X, sequence_length=seq_length, dtype=tf.float32)
X_batch = np.array([
# t = 0 t = 1
[[0, 1, 2], [9, 8, 7]], # instance 0
[[3, 4, 5], [0, 0, 0]], # instance 1
])
seq_length_batch = np.array([2, 1])
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
outputs_val, states_val = sess.run([outputs, states],
feed_dict={X: X_batch, seq_length: seq_length_batch})
print('outputs:')
print(outputs_val)
print('\nstates:')
print(states_val)
这会打印如下内容:
outputs:
[[[-0.85381496 -0.19517037 0.36011398 -0.18617202 0.39162001]
[-0.99998015 -0.99461144 -0.82241321 0.93778896 0.90737367]]
[[-0.99849552 -0.88643843 0.20635395 0.157896 0.76042926]
[ 0. 0. 0. 0. 0. ]]] # because len=1
states:
[[-0.99998015 -0.99461144 -0.82241321 0.93778896 0.90737367]
[-0.99849552 -0.88643843 0.20635395 0.157896 0.76042926]]
请注意,states
拥有与中相同的向量output
,它们是每个批处理实例的最后一个非零输出。