Dequantize

public final class Dequantize

Takes the packed uint32 input and unpacks the input to uint8 to do

Dequantization on device.

Constants

String OP_NAME The name of this op, as known by TensorFlow core engine

Public Methods

Output<TBfloat16>
asOutput()
Returns the symbolic handle of the tensor.
static Dequantize
create(Scope scope, Operand<?> input, Float minRange, Float maxRange, String mode, Boolean transposeOutput)
Factory method to create a class wrapping a new Dequantize operation.
Output<TBfloat16>
output()
Output tensors whose types is bloat16.

Inherited Methods

Constants

public static final String OP_NAME

The name of this op, as known by TensorFlow core engine

Constant Value: "XlaDequantize"

Public Methods

public Output<TBfloat16> asOutput ()

Returns the symbolic handle of the tensor.

Inputs to TensorFlow operations are outputs of another TensorFlow operation. This method is used to obtain a symbolic handle that represents the computation of the input.

public static Dequantize create (Scope scope, Operand<?> input, Float minRange, Float maxRange, String mode, Boolean transposeOutput)

Factory method to create a class wrapping a new Dequantize operation.

Parameters
scope current scope
input Input tensors whose types is uint32, shape is [d0, ..., dn].
minRange The minimum scalar value possibly produced for the input.
maxRange The maximum scalar value possibly produced for the input.
mode String to determine the dequantize mode in {"MIN_COMBINED", "MIN_FIRST", "SCALED"}.
transposeOutput Boolean to determine if output is transposed. transpose_output is faster when input is large and rank of input is higher than 1.
Returns
  • a new instance of Dequantize

public Output<TBfloat16> output ()

Output tensors whose types is bloat16. If transpose_output is true, output shape is [dn * 4, dn-1, ..., d1, d0]. If transpose_output is false, output shape is [d0,..., dn * 4].