Save the date! Google I/O returns May 18-20 Register now


Applies softmax to a batched N-D SparseTensor.

The inputs represent an N-D SparseTensor with logical shape [..., B, C] (where N >= 2), and with indices sorted in the canonical lexicographic order.

This op is equivalent to applying the normal tf.nn.softmax() to each innermost logical submatrix with shape [B, C], but with the catch that the implicitly zero elements do not participate. Specifically, the algorithm is equivalent to the following:

(1) Applies tf.nn.softmax() to a densified view of each innermost submatrix with shape [B, C], along the size-C dimension; (2) Masks out the original implicitly-zero locations; (3) Renormalizes the remaining elements.

Hence, the SparseTensor result has exactly the same non-zero indices and shape.

sp_indices A Tensor of type int64. 2-D. NNZ x R matrix with the indices of non-empty values in a SparseTensor, in canonical ordering.
sp_values A Tensor. Must be one of the following types: float32, float64. 1-D. NNZ non-empty values corresponding to sp_indices.
sp_shape A Tensor of type int64. 1-D. Shape of the input SparseTensor.
name A name for the operation (optional).

A Tensor. Has the same type as sp_values.