tf.keras.layers.RNN vs tf.keras.layers.StackedRNNCells: Tensorflow 2

Question

我正在尝试在Tensorflow 2.0中实现一个多层RNN模型。同时尝试tf.keras.layers.StackedRNNCells和tf.keras.layers.RNN会得到相同的结果。有人能帮我理解一下tf.keras.layers.RNN和tf.keras.layers.StackedRNNCells之间的区别吗

# driving parameters
sz_batch = 128
sz_latent = 200
sz_sequence = 196
sz_feature = 2
n_units = 120
n_layers = 3

带tf.keras.layers.RNN的多层RNN

inputs = tf.keras.layers.Input(batch_shape=(sz_batch, sz_sequence, sz_feature))
cells = [tf.keras.layers.GRUCell(n_units) for _ in range(n_layers)]
outputs = tf.keras.layers.RNN(cells, stateful=True, return_sequences=True, return_state=False)(inputs)
outputs = tf.keras.layers.Dense(1)(outputs)
model = tf.keras.Model(inputs=inputs, outputs=outputs)
model.summary()

返回：

Model: "model_13"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_88 (InputLayer)        [(128, 196, 2)]           0         
_________________________________________________________________
rnn_61 (RNN)                 (128, 196, 120)           218880    
_________________________________________________________________
dense_19 (Dense)             (128, 196, 1)             121       
=================================================================
Total params: 219,001
Trainable params: 219,001
Non-trainable params: 0

使用tf.keras.layers.RNN和tf.keras.layers.StackedRNNCells的多层RNN

inputs = tf.keras.layers.Input(batch_shape=(sz_batch, sz_sequence, sz_feature))
cells = [tf.keras.layers.GRUCell(n_units) for _ in range(n_layers)]
outputs = tf.keras.layers.RNN(tf.keras.layers.StackedRNNCells(cells),
                              stateful=True, 
                              return_sequences=True, 
                              return_state=False)(inputs)
outputs = tf.keras.layers.Dense(1)(outputs)
model = tf.keras.Model(inputs=inputs, outputs=outputs)
model.summary()

返回：

Model: "model_14"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_89 (InputLayer)        [(128, 196, 2)]           0         
_________________________________________________________________
rnn_62 (RNN)                 (128, 196, 120)           218880    
_________________________________________________________________
dense_20 (Dense)             (128, 196, 1)             121       
=================================================================
Total params: 219,001
Trainable params: 219,001
Non-trainable params: 0

Answer 1

如果您为tf.keras.layers.RNN提供一个列表或单元元组，它将使用tf.keras.layers.StackedRNNCells。这是在https://github.com/tensorflow/tensorflow/blob/v2.1.0/tensorflow/python/keras/layers/recurrent.py#L390中完成的