在自定义网络层时,training参数的作用是什么?
def call(self, inputs, training=None):
x = inputs
# embedding [b,80]=>[b,80,q00]
x = self.embedding(x)
state0 = self.state0
for word in tf.unstack(x, axis=1): # word:[b,100]
# h = tf.zeros(unit,)
out0, state0 = self.rnn_cell0(word, state0, training)
x = self.rnn_fc1(out0)
prob = tf.sigmoid(x)
return prob