How to return list of sparse tensor in Dataset?

TensorFlow General Discussion
, ,
1

I need to return sparse tensor and tensor in a Dataset, it looks like this:

import tensorflow as tf

def parse(x):
    s1 = tf.sparse.from_dense([0, 0, 2, 1])
    s2 = tf.sparse.from_dense([1, 0])
    return [s1, s2], tf.constant([1, 2, 3])


dataset = tf.data.Dataset.from_tensor_slices(list(range(10)))
dataset = dataset.batch(2)
dataset = dataset.map(parse)

for d in dataset.take(1):
    print(d)

But error occur:

TypeError: Neither a SparseTensor nor SparseTensorValue: [<tensorflow.python.framework.sparse_tensor.SparseTensor object at 0x7f1c5089cc10>, <tensorflow.python.framework.sparse_tensor.SparseTensor object at 0x7f1c9ca4b4c0>].

If return only [s1, s2], it works well.

My question is how to return list of sparse Tensor along with tensor

3

You should include Tensor into sparse_tensor list.
Working code snippet

import tensorflow as tf

def parse(x):
    s1 = tf.sparse.from_dense([0, 0, 2, 1])
    s2 = tf.sparse.from_dense([0, 1])
    return [s1, s2, tf.constant([1, 2, 3])]

dataset = tf.data.Dataset.from_tensor_slices(list(range(10)))
dataset = dataset.batch(2)
dataset = dataset.map(parse)
for d in dataset.take(1):
    print(d)