python静态函数_python 静态函数

最近在学python中发现，在类中声明为静态函数的函数在类内调用时可以不用跟上参数

例如这一段的self.msr_error函数声明为静态，compute_cost中调用此函数时不必声明参数，直接用self.msr_error就能调用

 def compute_cost(self):
        #losses = tf.losses.mean_squared_error(labels=self.ys, predictions=self.pred)
        #求出的是每一步的loss
        tf.reshape(self.pred,[-1,1])
        tf.reshape(self.ys,[-1,1])
        losses = tf.contrib.legacy_seq2seq.sequence_loss_by_example(
            [tf.reshape(self.pred, [-1], name='reshape_pred')],
            [tf.reshape(self.ys, [-1], name='reshape_target')],
            [tf.ones([self.batch_size * self.n_steps], dtype=tf.float32)],
            average_across_timesteps=True,
            softmax_loss_function=self.msr_error,
            name='losses'
        )
        # 计算batch_size内的平均的loss的值
        with tf.name_scope('average_cost'):
            self.cost = tf.div(
                tf.reduce_sum(losses, name='losses_num'),
                tf.cast(self.batch_size, tf.float32),
                name='average_cost')
            tf.summary.scalar('cost', self.cost)

    @staticmethod
    def msr_error(labels, logits):
        return tf.square(tf.subtract(labels, logits))
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24