ValueError: Can't convert non-rectangular Python sequence to Tensor

Question

I want to change list to tensor with tf.convert_to_tensor, data is following:

data=[

array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   1., 0., 0.]), 
array([0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.]), 
array([0., 0., 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
   0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.,
   0., 0., 0.]), 
array([0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0.])
    ]

it didn't work, system says:

ValueError: Can't convert non-rectangular Python sequence to Tensor.

how to solve this problem?

score 35 · Answer 1 · edited Apr 29 '22 at 10:24

35

I'm not sure whether they exist in TensorFlow 1 but TensorFlow 2.0 supports RaggedTensors, which the documentation describes as "... the TensorFlow equivalent of nested variable-length lists."

I think it would be trivial to convert your data to RaggedTensors. It might even be as easy as:

data_tensor = tf.ragged.constant(data)

Example:

>>> a = tf.ragged.constant([[1],[2,3]])
>>> a
<tf.RaggedTensor [[1], [2, 3]]>

edited Apr 29 '22 at 10:24

Maifee Ul Asad

3,992
6
38
86

answered Nov 04 '19 at 13:34

Eric McLachlan

3,132
2
25
37

1

ragged tensor indeed will add padding zeros? – Dee Apr 19 '21 at 07:59
1

Not by default, no. Check out the documentation: https://www.tensorflow.org/guide/ragged_tensor – Eric McLachlan Apr 19 '21 at 11:18

score 10 · Answer 2 · answered May 25 '19 at 15:59

10

You can't. Like the error message says, TensorFlow arrays can not have different sizes along one dimension. Try to use a list of TensorFlow arrays instead or the dataset api.

answered May 25 '19 at 15:59

BlueSun

3,541
1
18
37

Here is one example I gave of using the tf.data.Dataset API, but the tensors were all the same shape: https://stackoverflow.com/a/66175008/12763497 – brethvoice Feb 12 '21 at 15:56

ValueError: Can't convert non-rectangular Python sequence to Tensor

2 Answers2

Linked