Deprecated!!!, keeping it as Katka Zmolikova used it in JSALT 2015. More...

#include <nnet-sentence-averaging-component.h>

Inheritance diagram for SentenceAveragingComponent:

Collaboration diagram for SentenceAveragingComponent:

[legend]

Public Member Functions
	SentenceAveragingComponent (int32 dim_in, int32 dim_out)

	~SentenceAveragingComponent ()

Component *	Copy () const
	Copy component (deep copy),. More...

ComponentType	GetType () const
	Get Type Identification of the component,. More...

void	InitData (std::istream &is)
	Initialize the content of the component by the 'line' from the prototype,. More...

void	ReadData (std::istream &is, bool binary)
	Reads the component content. More...

void	WriteData (std::ostream &os, bool binary) const
	Writes the component content. More...

int32	NumParams () const
	Number of trainable parameters,. More...

void	GetGradient (VectorBase< BaseFloat > *gradient) const
	Get gradient reshaped as a vector,. More...

void	GetParams (VectorBase< BaseFloat > *params) const
	Get the trainable parameters reshaped as a vector,. More...

void	SetParams (const VectorBase< BaseFloat > &params)
	Set the trainable parameters from, reshaped as a vector,. More...

std::string	Info () const
	Print some additional info (after <ComponentName> and the dims),. More...

std::string	InfoGradient () const
	Print some additional info about gradient (after <...> and dims),. More...

void	PropagateFnc (const CuMatrixBase< BaseFloat > &in, CuMatrixBase< BaseFloat > *out)
	Abstract interface for propagation/backpropagation. More...

void	BackpropagateFnc (const CuMatrixBase< BaseFloat > &in, const CuMatrixBase< BaseFloat > &out, const CuMatrixBase< BaseFloat > &out_diff, CuMatrixBase< BaseFloat > *in_diff)
	Backward pass transformation (to be implemented by descending class...) More...

void	Update (const CuMatrixBase< BaseFloat > &input, const CuMatrixBase< BaseFloat > &diff)
	Compute gradient and update parameters,. More...

void	SetTrainOptions (const NnetTrainOptions &opts)
	Set the training options to the component,. More...

Public Member Functions inherited from UpdatableComponent
	UpdatableComponent (int32 input_dim, int32 output_dim)

virtual	~UpdatableComponent ()

bool	IsUpdatable () const
	Check if contains trainable parameters,. More...

const NnetTrainOptions &	GetTrainOptions () const
	Get the training options from the component,. More...

virtual void	SetLearnRateCoef (BaseFloat val)
	Set the learn-rate coefficient,. More...

virtual void	SetBiasLearnRateCoef (BaseFloat val)
	Set the learn-rate coefficient for bias,. More...

Public Member Functions inherited from Component
	Component (int32 input_dim, int32 output_dim)
	Generic interface of a component,. More...

virtual	~Component ()

virtual bool	IsMultistream () const
	Check if component has 'Recurrent' interface (trainable and recurrent),. More...

int32	InputDim () const
	Get the dimension of the input,. More...

int32	OutputDim () const
	Get the dimension of the output,. More...

void	Propagate (const CuMatrixBase< BaseFloat > &in, CuMatrix< BaseFloat > *out)
	Perform forward-pass propagation 'in' -> 'out',. More...

void	Backpropagate (const CuMatrixBase< BaseFloat > &in, const CuMatrixBase< BaseFloat > &out, const CuMatrixBase< BaseFloat > &out_diff, CuMatrix< BaseFloat > *in_diff)
	Perform backward-pass propagation 'out_diff' -> 'in_diff'. More...

void	Write (std::ostream &os, bool binary) const
	Write the component to a stream,. More...

Private Attributes
Nnet	nnet_

float	learn_rate_factor_

Additional Inherited Members
Public Types inherited from Component
enum	ComponentType { kUnknown = 0x0, kUpdatableComponent = 0x0100, kAffineTransform, kLinearTransform, kConvolutionalComponent, kLstmProjected, kBlstmProjected, kRecurrentComponent, kActivationFunction = 0x0200, kSoftmax, kHiddenSoftmax, kBlockSoftmax, kSigmoid, kTanh, kParametricRelu, kDropout, kLengthNormComponent, kTranform = 0x0400, kRbm, kSplice, kCopy, kTranspose, kBlockLinearity, kAddShift, kRescale, kKlHmm = 0x0800, kSentenceAveragingComponent, kSimpleSentenceAveragingComponent, kAveragePoolingComponent, kMaxPoolingComponent, kFramePoolingComponent, kParallelComponent, kMultiBasisComponent }
	Component type identification mechanism,. More...

Static Public Member Functions inherited from Component
static const char *	TypeToMarker (ComponentType t)
	Converts component type to marker,. More...

static ComponentType	MarkerToType (const std::string &s)
	Converts marker to component type (case insensitive),. More...

static Component *	Init (const std::string &conf_line)
	Initialize component from a line in config file,. More...

static Component *	Read (std::istream &is, bool binary)
	Read the component from a stream (static method),. More...

Static Public Attributes inherited from Component
static const struct key_value	kMarkerMap []
	The table with pairs of Component types and markers (defined in nnet-component.cc),. More...

Protected Attributes inherited from UpdatableComponent
NnetTrainOptions	opts_
	Option-class with training hyper-parameters,. More...

BaseFloat	learn_rate_coef_
	Scalar applied to learning rate for weight matrices (to be used in ::Update method),. More...

BaseFloat	bias_learn_rate_coef_
	Scalar applied to learning rate for bias (to be used in ::Update method),. More...

Protected Attributes inherited from Component
int32	input_dim_
	Data members,. More...

int32	output_dim_
	Dimension of the output of the Component,. More...

Detailed Description

Deprecated!!!, keeping it as Katka Zmolikova used it in JSALT 2015.

Definition at line 180 of file nnet-sentence-averaging-component.h.

Constructor & Destructor Documentation

◆ SentenceAveragingComponent()

SentenceAveragingComponent	(	int32	dim_in,
		int32	dim_out
	)

inline

Definition at line 182 of file nnet-sentence-averaging-component.h.

                                                          :
     UpdatableComponent(dim_in, dim_out), learn_rate_factor_(100.0)
   { }

◆ ~SentenceAveragingComponent()

~SentenceAveragingComponent ( )

inline

Definition at line 185 of file nnet-sentence-averaging-component.h.

186 { }

Member Function Documentation

◆ BackpropagateFnc()

void BackpropagateFnc	(	const CuMatrixBase< BaseFloat > &	in,
		const CuMatrixBase< BaseFloat > &	out,
		const CuMatrixBase< BaseFloat > &	out_diff,
		CuMatrixBase< BaseFloat > *	in_diff
	)

inlinevirtual

Backward pass transformation (to be implemented by descending class...)

Implements Component.

Definition at line 267 of file nnet-sentence-averaging-component.h.

References CuMatrixBase< Real >::ColRange(), CuMatrixBase< Real >::CopyFromMat(), and CuMatrixBase< Real >::NumCols().

                                                           {
     if (in_diff == NULL) return;
     int32 num_inputs = in.NumCols(),
       nnet_outputs = nnet_.OutputDim();
     in_diff->CopyFromMat(out_diff.ColRange(nnet_outputs, num_inputs));
   }

◆ Copy()

Component* Copy ( ) const

inlinevirtual

Copy component (deep copy),.

Implements Component.

Definition at line 188 of file nnet-sentence-averaging-component.h.

188 { return new SentenceAveragingComponent(*this); }

kaldi::nnet1::SentenceAveragingComponent::SentenceAveragingComponent

SentenceAveragingComponent(int32 dim_in, int32 dim_out)

Definition: nnet-sentence-averaging-component.h:182

◆ GetGradient()

void GetGradient ( VectorBase< BaseFloat > * gradient ) const

inlinevirtual

Get gradient reshaped as a vector,.

Implements UpdatableComponent.

Definition at line 225 of file nnet-sentence-averaging-component.h.

References KALDI_ERR.

                                                           {
     KALDI_ERR << "Unimplemented!";
   }

◆ GetParams()

void GetParams ( VectorBase< BaseFloat > * params ) const

inlinevirtual

Get the trainable parameters reshaped as a vector,.

Implements UpdatableComponent.

Definition at line 229 of file nnet-sentence-averaging-component.h.

References VectorBase< Real >::CopyFromVec(), VectorBase< Real >::Dim(), and KALDI_ASSERT.

                                                       {
     KALDI_ASSERT(params->Dim() == NumParams());
     Vector<BaseFloat> params_aux;
     nnet_.GetParams(&params_aux);
     params->CopyFromVec(params_aux);
   }

◆ GetType()

ComponentType GetType ( ) const

inlinevirtual

Get Type Identification of the component,.

Implements Component.

Definition at line 189 of file nnet-sentence-averaging-component.h.

References Component::kSentenceAveragingComponent.

189 { return kSentenceAveragingComponent; }

kaldi::nnet1::Component::kSentenceAveragingComponent

Definition: nnet-component.h:78

◆ Info()

std::string Info ( ) const

inlinevirtual

Print some additional info (after <ComponentName> and the dims),.

Reimplemented from Component.

Definition at line 240 of file nnet-sentence-averaging-component.h.

                          {
     return std::string("nested_network {\n") + nnet_.Info() + "}\n";
   }

◆ InfoGradient()

std::string InfoGradient ( ) const

inlinevirtual

Print some additional info about gradient (after <...> and dims),.

Reimplemented from Component.

Definition at line 244 of file nnet-sentence-averaging-component.h.

                                  {
     return std::string("nested_gradient {\n") + nnet_.InfoGradient() + "}\n";
   }

◆ InitData()

void InitData ( std::istream & is )

inlinevirtual

Initialize the content of the component by the 'line' from the prototype,.

Implements UpdatableComponent.

Definition at line 191 of file nnet-sentence-averaging-component.h.

References Component::InputDim(), KALDI_ASSERT, KALDI_ERR, Component::OutputDim(), kaldi::ReadBasicType(), and kaldi::ReadToken().

                                 {
     // define options
     std::string nested_nnet_filename;
     std::string nested_nnet_proto;
     // parse config
     std::string token;
     while (is >> std::ws, !is.eof()) {
       ReadToken(is, false, &token);
        if (token == "<NestedNnetFilename>") ReadToken(is, false, &nested_nnet_filename);
       else if (token == "<NestedNnetProto>") ReadToken(is, false, &nested_nnet_proto);
       else if (token == "<LearnRateFactor>") ReadBasicType(is, false, &learn_rate_factor_);
       else KALDI_ERR << "Unknown token " << token << " Typo in config?";
     }
     // initialize (read already prepared nnet from file)
     KALDI_ASSERT((nested_nnet_proto != "") ^ (nested_nnet_filename != ""));  // xor,
     if (nested_nnet_filename != "") nnet_.Read(nested_nnet_filename);
     if (nested_nnet_proto != "") nnet_.Init(nested_nnet_proto);
     // check dims of nested nnet
     KALDI_ASSERT(InputDim() == nnet_.InputDim());
     KALDI_ASSERT(OutputDim() == nnet_.OutputDim() + InputDim());
   }

◆ NumParams()

int32 NumParams ( ) const

inlinevirtual

Number of trainable parameters,.

Implements UpdatableComponent.

Definition at line 223 of file nnet-sentence-averaging-component.h.

223 { return nnet_.NumParams(); }

kaldi::nnet1::Nnet::NumParams

int32 NumParams() const

Get the number of parameters in the network,.

Definition: nnet-nnet.cc:210

kaldi::nnet1::SentenceAveragingComponent::nnet_

Nnet nnet_

Definition: nnet-sentence-averaging-component.h:306

◆ PropagateFnc()

void PropagateFnc	(	const CuMatrixBase< BaseFloat > &	in,
		CuMatrixBase< BaseFloat > *	out
	)

inlinevirtual

Abstract interface for propagation/backpropagation.

Forward pass transformation (to be implemented by descending class...)

Implements Component.

Definition at line 248 of file nnet-sentence-averaging-component.h.

References CuVectorBase< Real >::AddRowSumMat(), CuMatrixBase< Real >::ColRange(), CuMatrixBase< Real >::NumCols(), and CuMatrixBase< Real >::NumRows().

                                                   {
     // Get NN output
     CuMatrix<BaseFloat> out_nnet;
     nnet_.Propagate(in, &out_nnet);
     // Get the average row (averaging over the time axis):
     // averaging corresponds to extraction of a 'constant vector'
     // code for single sentence,
     int32 num_inputs = in.NumCols(),
       nnet_outputs = nnet_.OutputDim(),
       num_frames = out_nnet.NumRows();
 
     CuVector<BaseFloat> average_row(nnet_outputs);
     average_row.AddRowSumMat(1.0/num_frames, out_nnet, 0.0);
     // Forwarding sentence codes along with input features
     out->ColRange(0, nnet_outputs).AddVecToRows(1.0, average_row, 0.0);
     out->ColRange(nnet_outputs, num_inputs).CopyFromMat(in);
   }

◆ ReadData()

void ReadData	(	std::istream &	is,
		bool	binary
	)

inlinevirtual

Reads the component content.

Reimplemented from Component.

Definition at line 213 of file nnet-sentence-averaging-component.h.

References Component::InputDim(), KALDI_ASSERT, and Component::OutputDim().

                                              {
     nnet_.Read(is, binary);
     KALDI_ASSERT(nnet_.InputDim() == InputDim());
     KALDI_ASSERT(nnet_.OutputDim() + InputDim() == OutputDim());
   }

◆ SetParams()

void SetParams ( const VectorBase< BaseFloat > & params )

inlinevirtual

Set the trainable parameters from, reshaped as a vector,.

Implements UpdatableComponent.

Definition at line 236 of file nnet-sentence-averaging-component.h.

References KALDI_ERR.

                                                       {
     KALDI_ERR << "Unimplemented!";
   }

◆ SetTrainOptions()

void SetTrainOptions ( const NnetTrainOptions & opts )

inlinevirtual

Set the training options to the component,.

Reimplemented from UpdatableComponent.

Definition at line 297 of file nnet-sentence-averaging-component.h.

References NnetTrainOptions::learn_rate, and UpdatableComponent::SetTrainOptions().

                                                      {
     UpdatableComponent::SetTrainOptions(opts_);
     // Pass the train options to the nnet
     NnetTrainOptions o(opts);
     o.learn_rate *= learn_rate_factor_;
     nnet_.SetTrainOptions(opts_);
   }

◆ Update()

void Update	(	const CuMatrixBase< BaseFloat > &	input,
		const CuMatrixBase< BaseFloat > &	diff
	)

inlinevirtual

Compute gradient and update parameters,.

Implements UpdatableComponent.

Definition at line 277 of file nnet-sentence-averaging-component.h.

References CuVectorBase< Real >::AddRowSumMat(), CuMatrixBase< Real >::AddVecToRows(), CuMatrixBase< Real >::ColRange(), and CuMatrixBase< Real >::NumRows().

                                                    {
     // get useful dims,
     int32 nnet_outputs = nnet_.OutputDim(),
       num_frames = diff.NumRows();
     // Passing the derivative into the nested network. The loss derivative is averaged:
     // single frame from nested network influenced all the frames in the main network,
     // so to get the derivative w.r.t. single frame from nested network we sum derivatives
     // of all frames from main network (and scale by 1/Nframes constant).
     //
     // In fact all the frames from nested network influenced all the input frames to main nnet,
     // so the loss derivarive w.r.t. nested network output is same for all frames in sentence.
     CuVector<BaseFloat> average_diff(nnet_outputs);
     average_diff.AddRowSumMat(1.0 / num_frames, diff.ColRange(0, nnet_outputs), 0.0);
     CuMatrix<BaseFloat> nnet_out_diff(num_frames, nnet_outputs);
     nnet_out_diff.AddVecToRows(1.0, average_diff, 0.0);
     //
     nnet_.Backpropagate(nnet_out_diff, NULL);
   }

◆ WriteData()

void WriteData	(	std::ostream &	os,
		bool	binary
	)		const

inlinevirtual

Writes the component content.

Reimplemented from Component.

Definition at line 219 of file nnet-sentence-averaging-component.h.

                                                     {
     nnet_.Write(os, binary);
   }

Member Data Documentation

◆ learn_rate_factor_

float learn_rate_factor_

private

Definition at line 307 of file nnet-sentence-averaging-component.h.

◆ nnet_

Nnet nnet_

private

Definition at line 306 of file nnet-sentence-averaging-component.h.

The documentation for this class was generated from the following file:

nnet/nnet-sentence-averaging-component.h

Public Member Functions

Private Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ SentenceAveragingComponent()

◆ ~SentenceAveragingComponent()

Member Function Documentation

◆ BackpropagateFnc()

◆ Copy()

◆ GetGradient()

◆ GetParams()

◆ GetType()

◆ Info()

◆ InfoGradient()

◆ InitData()

◆ NumParams()

◆ PropagateFnc()

◆ ReadData()

◆ SetParams()

◆ SetTrainOptions()

◆ Update()

◆ WriteData()

Member Data Documentation

◆ learn_rate_factor_

◆ nnet_