Have a question? Connect with the community at the TensorFlow Forum Visit Forum

FakeQuantWithMinMaxVarsPerChannelGradient

public final class FakeQuantWithMinMaxVarsPerChannelGradient

Compute gradients for a FakeQuantWithMinMaxVarsPerChannel operation.

Nested Classes

class FakeQuantWithMinMaxVarsPerChannelGradient.Options Optional attributes for FakeQuantWithMinMaxVarsPerChannelGradient

Constants

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

Public Methods

Output < TFloat32 >
backpropWrtMax ()
Backpropagated gradients w.r.t.
Output < TFloat32 >
backpropWrtMin ()
Backpropagated gradients w.r.t.
Output < TFloat32 >
backpropsWrtInput ()
Backpropagated gradients w.r.t.
static FakeQuantWithMinMaxVarsPerChannelGradient
create ( Scope scope, Operand < TFloat32 > gradients, Operand < TFloat32 > inputs, Operand < TFloat32 > min, Operand < TFloat32 > max, Options... options)
Factory method to create a class wrapping a new FakeQuantWithMinMaxVarsPerChannelGradient operation.
static FakeQuantWithMinMaxVarsPerChannelGradient.Options
narrowRange (Boolean narrowRange)
static FakeQuantWithMinMaxVarsPerChannelGradient.Options
numBits (Long numBits)

Inherited Methods

Constants

public static final String OP_NAME

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

Constant Value: "FakeQuantWithMinMaxVarsPerChannelGradient"

Public Methods

public Output < TFloat32 > backpropWrtMax ()

Backpropagated gradients w.r.t. max parameter, shape `[d]`: `sum_per_d(gradients * (inputs > max))`.

public Output < TFloat32 > backpropWrtMin ()

Backpropagated gradients w.r.t. min parameter, shape `[d]`: `sum_per_d(gradients * (inputs < min))`.

public Output < TFloat32 > backpropsWrtInput ()

Backpropagated gradients w.r.t. inputs, shape same as `inputs`: `gradients * (inputs >= min && inputs <= max)`.

public static FakeQuantWithMinMaxVarsPerChannelGradient create ( Scope scope, Operand < TFloat32 > gradients, Operand < TFloat32 > inputs, Operand < TFloat32 > min, Operand < TFloat32 > max, Options... options)

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

Parameters
scope current scope
gradients Backpropagated gradients above the FakeQuantWithMinMaxVars operation, shape one of: `[d]`, `[b, d]`, `[b, h, w, d]`.
inputs Values passed as inputs to the FakeQuantWithMinMaxVars operation, shape same as `gradients`. min, max: Quantization interval, floats of shape `[d]`.
options carries optional attributes values
Returns
  • a new instance of FakeQuantWithMinMaxVarsPerChannelGradient

public static FakeQuantWithMinMaxVarsPerChannelGradient.Options narrowRange (Boolean narrowRange)

Parameters
narrowRange Whether to quantize into 2^num_bits - 1 distinct values.

public static FakeQuantWithMinMaxVarsPerChannelGradient.Options numBits (Long numBits)

Parameters
numBits The bitwidth of the quantization; between 2 and 16, inclusive.