com.linkedin.feathr.common.tensor

Interface TensorData

All Known Implementing Classes:

ByteBufferDenseTensor, CoercingTensorData, DenseBooleanTensor, DenseBytesTensor, DenseDoubleTensor, DenseFloatTensor, DenseIntTensor, DenseLongTensor, DenseStringTensor, DenseTensor, LOLTensorData, ScalarBooleanTensor, ScalarBytesTensor, ScalarDoubleTensor, ScalarFloatTensor, ScalarIntTensor, ScalarLongTensor, ScalarStringTensor, ScalarTensor, UniversalTensor

public interface TensorData

A typed container of structured data. Maps a (potentially empty) tuple of dimensions to a scalar value. For efficiency, a single object represents the entire data structure. Sequential access is provided via TensorIterator. The typical use of TensorData is:

 TensorData td = ...;
 TensorIterator ti = td.iterator();
 ti.start();
 while (ti.isValid()) {
  ti.getInt(0);
  ti.getFloat(1);
  ti.getInt(2);
 // and others getXXX(),
 // the getXXX(dimensionality) corresponding to the value
  ti.next();
 }
 

Method Summary

Modifier and Type	Method and Description
int	cardinality()
int	estimatedCardinality()
default int	getArity()
default long[]	getShape()
Representable[]	getTypes()
default boolean	isEmpty()
TensorIterator	iterator()

Method Detail

getArity

default int getArity()

Returns:

the arity, which is the total number of columns. As columns are dimensions plus a value, arity is by one greater than dimensionality.

getTypes

Representable[] getTypes()

Returns:

the types of all dimensions and the value as a single array.

estimatedCardinality

int estimatedCardinality()

Returns:

conservative estimate of number of entries in the mapping from dimensions to values. Can be cheaper to evaluate than the exact one.

cardinality

int cardinality()

Returns:

exact number of entries in the mapping from dimensions to values.

isEmpty

default boolean isEmpty()

iterator

TensorIterator iterator()

Returns:

an iterator over entries in the mapping from dimensions to the value.

getShape

default long[] getShape()

Returns:

the current shape of the data or the supplied shape of the data