我构建了一个变分自动编码器，它具有潜在分布作为多元正态分布。我想存储分发并将其可视化（在tensorboard或其他软件中）。但是，当我尝试使用add_summary中的FileWriter时，我收到错误 - MultivariateNormalDiag object has no attribute 'value'。如何存储和显示高斯分布？
代码：
import tensorflow as tf
tfd = tf.contrib.distribution
input = tf.placeholder(tf.float32, [None, 9])
x = tf.layers.dense(input,200, tf.nn.relu)
x = tf.layers.dense(x,200,tf.nn.relu)
loc = tf.layers.dense(x,1)
scale = tf.layers.dense(x,1,tf.nn.softplus)
latent = tfd.MultivariateNormalDiag(loc,scale)
# getting loss and minimizing it
writer = tf.summary.FileWriter("/tmp/dist")
writer.add_summary(latent) # Error over here