Python 解释为什么Tensorflow会抛出“；试图使用未初始化的值“；在线性变换过程中

在运行jupyter笔记本时，我试图在张量上应用

tf.layers.dense

。我正在使用的代码引发了失败的预处理错误：

FailedPreconditionError: Attempting to use uninitialized value dense_9/bias
     [[Node: dense_9/bias/read = Identity[T=DT_INT32, _class=["loc:@dense_9/bias"], _device="/job:localhost/replica:0/task:0/device:CPU:0"](dense_9/bias)]]

Jupyter笔记本代码：

import tensorflow as tf
import numpy as np
tf.InteractiveSession()
batch_size = 3
seqlen = 5
nfeats = 3
mini_batch = np.random.randint(1,10, (batch_size, seqlen, nfeats))
d = dict((n, t.eval()) for n,t in enumerate(tf.unstack(mini_batch, 3)))
d0 = tf.convert_to_tensor(d[0],dtype=tf.int32)
d0 = tf.reshape(d0, [seqlen, batch_size])
d0_lin = tf.layers.dense(inputs=d0, units=100)
d0_lin.eval()

---

我不熟悉Tensorflow，我正在玩的想法是如何对

5x3

张量应用线性变换，并在有

3x5x3

输入张量时将其转换为

5x100

张量。因此，我们可以将

3x5x3

张量转换为

3x5x100

张量。

因为您得到了一个

失败的预处理错误：试图使用未初始化的值，我初始化了全局变量，使其工作

import tensorflow as tf
import numpy as np
tf.InteractiveSession()
batch_size = 3
seqlen = 5
nfeats = 3
mini_batch = np.random.randint(1,10, (batch_size, seqlen, nfeats))
d = dict((n, t.eval()) for n,t in enumerate(tf.unstack(mini_batch, 3)))
d0 = tf.convert_to_tensor(d[0],dtype=tf.int32)
d0 = tf.reshape(d0, [seqlen, batch_size])
d0_lin = tf.layers.dense(inputs=d0, units=100)

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    print(d0_lin.eval())

输出：

[[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]]