Tensorflow: fill tensor inside a function

Question

What I'm trying to define is the following idea:

Consider whe have these tensors

a = tf.constant([1., 1.5, 1.2]) # tensor with shape [3,]
b = tf.constant([1., 2., 3.])   # ""
c = tf.constant([3., 0., 6.])   # ""

t = tf.constant([0.5, 0.6, 0.7, 2., 4., 5., 6.]) # tensor with shape [7,]

Now let's consider I want to compute a new tensor, working with each element of previous tensors, for example:

def new_tensor(a, b, c, t):
    X = tf.constant([[tf.sin(a*t[1]), b*t[3], c+t[4]], 
        [tf.cos(b*t[5]), tf.atan2(t[5], c), a+t[2]+b],  
        [a+t[4], a+b, c*t[0]]])
    return X

X should be a tensor with shape [3, 3, 3]. That is, I want to define a function that takes four tensors as input: three of them of same shape and the fourth with a different one. I want the function to compute a tensor (X) for each value of first three inputs (a, b, c).

With this code TensorFlow gives this error:

TypeError: List of Tensors when single Tensor expected

According to this post, this is beacuse tf.constant cannot take a tensor as input, and they recommend to use tf.Variable instead. But I don't think that fits for me because I have to work later with X and don't want to initializate it, etc. I have also read this other post, but couldn't find any answer to my problem.

Is there any way to do what I want? Does my code make sense for my purpose? Thank you in advance.

UPDATE: with jdehesa answer

Taking @jdehesa answer and making resulting tensor more simple:

def new_tensor(a, b, c, t):
    # Could also use tf.convert_to_tensor
    X = tf.stack([[a+t[1],   b*t[1],    c+t[1]],
                  [b*t[0],  t[5]+ c,  a+t[2]+b],
                  [a+t[4],      a+b,   c*t[0]]])
    return X

And with tensors:

a = tf.constant([1., 1., 1.]) # tensor with shape [3,]
b = tf.constant([2., 2., 2.])   # ""
c = tf.constant([3., 3., 3.])   # ""

t = tf.constant([1., 1., 1., 1., 1., 1., 1.]) # tensor with shape [7,]

What I get is the following tensor:

# When evaluating x = new_tensor(a,b,c,t)
[[[2. 2. 2.]
  [2. 2. 2.]
  [4. 4. 4.]]

 [[2. 2. 2.]
  [4. 4. 4.]
  [4. 4. 4.]]

 [[2. 2. 2.]
  [3. 3. 3.]
  [3. 3. 3.]]]

But what I would expected is the following:

[[[2. 2. 4.]
  [2. 4. 4.]
  [2. 3. 3.]]

 [[2. 2. 4.]
  [2. 4. 4.]
  [2. 3. 3.]]

 [[2. 2. 4.]
  [2. 4. 4.]
  [2. 3. 3.]]]

As what I want it to evaluate to each element of input tensors.

Answer 1

That's correct, you can only pass Python or NumPy values to tf.constant, but you can build your tensor with tf.stack or, if you prefer, generally with tf.convert_to_tensor:

import tensorflow as tf

def new_tensor(a, b, c, t):
    # Could also use tf.convert_to_tensor
    X = tf.stack([[tf.sin(a*t[1]),            b*t[3],   c+t[4]],
                  [tf.cos(b*t[5]), tf.atan2(t[5], c), a+t[2]+b],
                  [        a+t[4],               a+b,   c*t[0]]])
    return X

with tf.Graph().as_default(), tf.Session() as sess:
    a = tf.constant([1., 1.5, 1.2]) # tensor with shape [3,]
    b = tf.constant([1., 2., 3.])   # ""
    c = tf.constant([3., 0., 6.])   # ""
    t = tf.constant([0.5, 0.6, 0.7, 2., 4., 5., 6.]) # tensor with shape [7,]
    x = new_tensor(a, b, c, t)
    print(sess.run(x))
    # [[[ 0.5646425   0.7833269   0.65938467]
    #   [ 2.          4.          6.        ]
    #   [ 7.          4.         10.        ]]
    # 
    #  [[ 0.2836622  -0.8390715  -0.7596879 ]
    #   [ 1.0303768   1.5707964   0.69473827]
    #   [ 2.7         4.2         4.9       ]]
    # 
    #  [[ 5.          5.5         5.2       ]
    #   [ 2.          3.5         4.2       ]
    #   [ 1.5         0.          3.        ]]]

---

EDIT: For your second example, to get the result that you want you need to use tf.transpose to change the order of the dimensions of the tensor:

import tensorflow as tf

def new_tensor(a, b, c, t):
    # Could also use tf.convert_to_tensor
    X = tf.stack([[a+t[1],   b*t[1],    c+t[1]],
                  [b*t[0],  t[5]+ c,  a+t[2]+b],
                  [a+t[4],      a+b,   c*t[0]]])
    X = tf.transpose(X, (2, 0, 1))
    return X

with tf.Graph().as_default(), tf.Session() as sess:
    a = tf.constant([1., 1., 1.]) # tensor with shape [3,]
    b = tf.constant([2., 2., 2.])   # ""
    c = tf.constant([3., 3., 3.])   # ""
    t = tf.constant([1., 1., 1., 1., 1., 1., 1.]) # tensor with shape [7,]
    x = new_tensor(a, b, c, t)
    print(sess.run(x))
    # [[[2. 2. 4.]
    #   [2. 4. 4.]
    #   [2. 3. 3.]]
    # 
    #  [[2. 2. 4.]
    #   [2. 4. 4.]
    #   [2. 3. 3.]]
    # 
    #  [[2. 2. 4.]
    #   [2. 4. 4.]
    #   [2. 3. 3.]]]