TypeError: List of Tensors when single Tensor expected - when using const with tf.random_normal

Question

I have the following TensorFlow code:

tf.constant(tf.random_normal([time_step, batch_size], -1, 1))

I am getting TypeError: List of Tensors when single Tensor expected. Could you tell me what is wrong with the code?

Answer 1

Someone else has answered this question on another thread.

Essentially, tf.constant() takes a NumPy array as an argument or some sort of array or just a value.

tf.random_normal() returns a Tensor which cannot be an argument to tf.constant().

To fix this, use tf.Variable() instead of tf.constant().

See the answer from the link. The person explains it better.

Answer 2

tf.constant is supposed to have a constant argument - value. Here value can be a constant value or a list of values of type dtype. You can't create a constant tensor that has another tensor as its value.