请问各位大佬，在这个注意力机制中的ratio和num分别代表啥

```def se_layer(inputs_tensor=None,ratio=None,num=None):

channels = K.int_shape(inputs_tensor)[-1]
x = tf.layers.AveragePooling1D()(inputs_tensor)
x = tf.layers.Reshape((1, 1, channels))(x)
x = tf.layers.Conv1D(channels//ratio, (1, 1), strides=1, name="se_conv1_"+str(num), padding="valid")(x)
x = tf.layers.Activation('relu', name='se_conv1_relu_'+str(num))(x)
x = tf.layers.Conv1D(channels, (1, 1), strides=1, name="se_conv2_"+str(num), padding="valid")(x)
x = tf.layers.Activation('sigmoid', name='se_conv2_relu_'+str(num))(x)
output = tf.layers.multiply([inputs_tensor, x])
return output

(channels//ratio这个是你的数据集读取的。从字面理解是一个比率。num是数量，具体要看模型和数据集。