Is there a way to reshape a tensor and pad any overflow with zeros? I know ndarray.reshape does this, but as I understand it, converting a Tensor to an ndarray would require flip-flopping between the GPU and CPU.
Tensorflow's reshape() documentation says the TensorShapes need to have the same number of elements, so perhaps the best way would be a pad() and then reshape()?
I'm trying to achieve:
a = tf.Tensor([[1,2],[3,4]])
tf.reshape(a, [2,3])
a => [[1, 2, 3],
[4, 0 ,0]]
Tensorflow now offers the pad function which performs padding on a tensor in a number of ways(like opencv2's padding function for arrays): https://www.tensorflow.org/api_docs/python/tf/pad
tf.pad(tensor, paddings, mode='CONSTANT', name=None)
example from the docs above:
# 't' is [[1, 2, 3], [4, 5, 6]].
# 'paddings' is [[1, 1,], [2, 2]].
# rank of 't' is 2.
pad(t, paddings, "CONSTANT") ==> [[0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 2, 3, 0, 0],
[0, 0, 4, 5, 6, 0, 0],
[0, 0, 0, 0, 0, 0, 0]]
pad(t, paddings, "REFLECT") ==> [[6, 5, 4, 5, 6, 5, 4],
[3, 2, 1, 2, 3, 2, 1],
[6, 5, 4, 5, 6, 5, 4],
[3, 2, 1, 2, 3, 2, 1]]
pad(t, paddings, "SYMMETRIC") ==> [[2, 1, 1, 2, 3, 3, 2],
[2, 1, 1, 2, 3, 3, 2],
[5, 4, 4, 5, 6, 6, 5],
[5, 4, 4, 5, 6, 6, 5]]
As far as I know, there's no built-in operator that does this (tf.reshape() will give you an error if the shapes do not match). However, you can achieve the same result with a few different operators:
a = tf.constant([[1, 2], [3, 4]])
# Reshape `a` as a vector. -1 means "set this dimension automatically".
a_as_vector = tf.reshape(a, [-1])
# Create another vector containing zeroes to pad `a` to (2 * 3) elements.
zero_padding = tf.zeros([2 * 3] - tf.shape(a_as_vector), dtype=a.dtype)
# Concatenate `a_as_vector` with the padding.
a_padded = tf.concat([a_as_vector, zero_padding], 0)
# Reshape the padded vector to the desired shape.
result = tf.reshape(a_padded, [2, 3])
