Search code examples
pythontensorflowsparse-matrix

tensorflow: Error multiplying a sparse matrix with a dense matrix using tf.matmul

In the following code, I want dense matrix B to left multiply a sparse matrix A, but I got errors.

import tensorflow as tf
import numpy as np

A = tf.sparse_placeholder(tf.float32)
B = tf.placeholder(tf.float32, shape=(5,5))
C = tf.matmul(B,A,a_is_sparse=False,b_is_sparse=True)
sess = tf.InteractiveSession()
indices = np.array([[3, 2], [1, 2]], dtype=np.int64)
values = np.array([1.0, 2.0], dtype=np.float32)
shape = np.array([5,5], dtype=np.int64)
Sparse_A = tf.SparseTensorValue(indices, values, shape)
RandB = np.ones((5, 5))
print sess.run(C, feed_dict={A: Sparse_A, B: RandB})

The error message is as follows:

TypeError: Failed to convert object of type <class 'tensorflow.python.framework.sparse_tensor.SparseTensor'> 
to Tensor. Contents: SparseTensor(indices=Tensor("Placeholder_4:0", shape=(?, ?), dtype=int64), values=Tensor("Placeholder_3:0", shape=(?,), dtype=float32), dense_shape=Tensor("Placeholder_2:0", shape=(?,), dtype=int64)). 
Consider casting elements to a supported type.

What's wrong with my code?

I'm doing this following the documentation and it says we should use a_is_sparse to denote whether the first matrix is sparse, and similarly with b_is_sparse. Why is my code wrong?

As is suggested by vijay, I should use C = tf.matmul(B,tf.sparse_tensor_to_dense(A),a_is_sparse=False,b_is_sparse=True)

I tried this but I met with another error saying:

Caused by op u'SparseToDense', defined at:
  File "a.py", line 19, in <module>
    C = tf.matmul(B,tf.sparse_tensor_to_dense(A),a_is_sparse=False,b_is_sparse=True)
  File "/home/mypath/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/sparse_ops.py", line 845, in sparse_tensor_to_dense
    name=name)
  File "/home/mypath/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/sparse_ops.py", line 710, in sparse_to_dense
    name=name)
  File "/home/mypath/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/gen_sparse_ops.py", line 1094, in _sparse_to_dense
    validate_indices=validate_indices, name=name)
  File "/home/mypath/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 767, in apply_op
    op_def=op_def)
  File "/home/mypath/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2506, in create_op
    original_op=self._default_original_op, op_def=op_def)
  File "/home/mypath/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1269, in __init__
    self._traceback = _extract_stack()

InvalidArgumentError (see above for traceback): indices[1] = [1,2] is out of order
[[Node: SparseToDense = SparseToDense[T=DT_FLOAT, Tindices=DT_INT64, validate_indices=true, _device="/job:localhost/replica:0/task:0/cpu:0"](_arg_Placeholder_4_0_2, _arg_Placeholder_2_0_0, _arg_Placeholder_3_0_1, SparseToDense/default_value)]]

Thank you all for helping me!

Solution

  • In tf.matmul, flags a_is_sparse and b_is_sparse do not indicate the operands being SparseTensors, but instead, they are algorithmic hints to invoke more efficient ways to compute multiplication on two dense Tensors. In your code should be :

    C = tf.matmul(B,tf.sparse_tensor_to_dense(A),a_is_sparse=False,b_is_sparse=True)

    To matmul a SparseTensor and a dense Tensor, you can also use tf.sparse_tensor_dense_matmul() instead.