ChunkInfo Class Reference

ChunkInfo is a class whose purpose is to describe the structure of matrices holding features. More...

#include <nnet-component.h>

Collaboration diagram for ChunkInfo:

Public Member Functions
	ChunkInfo ()
	ChunkInfo (int32 feat_dim, int32 num_chunks, int32 first_offset, int32 last_offset)
	ChunkInfo (int32 feat_dim, int32 num_chunks, const std::vector< int32 > offsets)
int32	GetIndex (int32 offset) const
int32	GetOffset (int32 index) const
void	MakeOffsetsContiguous ()
int32	ChunkSize () const
int32	NumChunks () const
int32	NumRows () const
	Returns the number of rows that we expect the feature matrix to have. More...
int32	NumCols () const
	Returns the number of columns that we expect the feature matrix to have. More...
void	CheckSize (const CuMatrixBase< BaseFloat > &mat) const
	Checks that the matrix has the size we expect, and die if not. More...
void	Check () const
	Checks that the data in the ChunkInfo is valid, and die if not. More...

Private Attributes
int32	feat_dim_
int32	num_chunks_
int32	first_offset_
int32	last_offset_
std::vector< int32 >	offsets_

Detailed Description

ChunkInfo is a class whose purpose is to describe the structure of matrices holding features.

This is useful mostly in training time. The main reason why we have this is to support efficient training for networks which we have splicing components that splice in a non-contiguous way, e.g. frames -5, 0 and 5. We also have in mind future extensibility to convnets which might have similar issues. This class describes the structure of a minibatch of features, or of a single contiguous block of features. Examples are as follows, and offsets is empty if not mentioned: When decoding, at input to the network: feat_dim = 13, num_chunks = 1, first_offset = 0, last_offset = 691 and in the middle of the network (assuming splicing is +-7): feat_dim = 1024, num_chunks = 1, first_offset = 7, last_offset = 684 When training, at input to the network: feat_dim = 13, num_chunks = 512, first_offset = 0, last_offset= 14 and in the middle of the network: feat_dim = 1024, num_chunks = 512, first_offset = 7, last_offset = 7 The only situation where offsets would be nonempty would be if we do splicing with gaps in. E.g. suppose at network input we splice +-2 frames (contiguous) and somewhere in the middle we splice frames {-5, 0, 5}, then we would have the following while training At input to the network: feat_dim = 13, num_chunks = 512, first_offset = 0, last_offset = 14 After the first hidden layer: feat_dim = 1024, num_chunks = 512, first_offset = 2, last_offset = 12, offsets = {2, 10, 12} At the output of the last hidden layer (after the {-5, 0, 5} splice): feat_dim = 1024, num_chunks = 512, first_offset = 7, last_offset = 7 (the decoding setup would still look pretty normal, so we don't give an example).

Definition at line 72 of file nnet-component.h.

Constructor & Destructor Documentation

ChunkInfo() [1/3]

ChunkInfo ( )

inline

Definition at line 74 of file nnet-component.h.

      : feat_dim_(0), num_chunks_(0),
        first_offset_(0), last_offset_(0),
        offsets_() { }

ChunkInfo() [2/3]

ChunkInfo ( int32  feat_dim, int32  num_chunks, int32  first_offset, int32  last_offset  )

inline

Definition at line 79 of file nnet-component.h.

References ChunkInfo::Check().

      : feat_dim_(feat_dim), num_chunks_(num_chunks),
        first_offset_(first_offset), last_offset_(last_offset),
        offsets_() { Check(); }

ChunkInfo() [3/3]

ChunkInfo ( int32  feat_dim, int32  num_chunks, const std::vector< int32 >  offsets  )

inline

Definition at line 85 of file nnet-component.h.

References ChunkInfo::Check(), ChunkInfo::first_offset_, ChunkInfo::GetIndex(), ChunkInfo::GetOffset(), ChunkInfo::last_offset_, and ChunkInfo::offsets_.

      : feat_dim_(feat_dim), num_chunks_(num_chunks),
        first_offset_(offsets.front()), last_offset_(offsets.back()),
        offsets_(offsets) { if (last_offset_ - first_offset_ + 1 == offsets_.size())
                              offsets_.clear();
          Check(); }

Member Function Documentation

Check()

void Check

(

)

const

Checks that the data in the ChunkInfo is valid, and die if not.

Definition at line 2543 of file nnet-component.cc.

References KALDI_ASSERT.

Referenced by SpliceComponent::Backprop(), SpliceMaxComponent::Backprop(), ChunkInfo::ChunkInfo(), ChunkInfo::MakeOffsetsContiguous(), ChunkInfo::NumCols(), SpliceComponent::Propagate(), and SpliceMaxComponent::Propagate().

                            {
  // Checking sanity of the ChunkInfo object
  KALDI_ASSERT((feat_dim_ > 0) && (num_chunks_ > 0));

  if (! offsets_.empty()) {
    KALDI_ASSERT((first_offset_ == offsets_.front()) &&
                 (last_offset_ == offsets_.back()));
  } else  {
    KALDI_ASSERT((first_offset_ >= 0) && (last_offset_ >= first_offset_));
    // asserting the chunk is not contiguous, as offsets is not empty
    KALDI_ASSERT ( last_offset_ - first_offset_ + 1 > offsets_.size() );
  }
  KALDI_ASSERT(NumRows() % num_chunks_ == 0);

}

CheckSize()

void CheckSize

(

const CuMatrixBase< BaseFloat > &

mat

)

const

Checks that the matrix has the size we expect, and die if not.

Definition at line 2559 of file nnet-component.cc.

References KALDI_ASSERT, CuMatrixBase< Real >::NumCols(), and CuMatrixBase< Real >::NumRows().

Referenced by SpliceComponent::Backprop(), SpliceMaxComponent::Backprop(), ChunkInfo::NumCols(), MaxoutComponent::Propagate(), MaxpoolingComponent::Propagate(), PnormComponent::Propagate(), SigmoidComponent::Propagate(), TanhComponent::Propagate(), PowerComponent::Propagate(), SoftHingeComponent::Propagate(), SoftmaxComponent::Propagate(), LogSoftmaxComponent::Propagate(), AffineComponent::Propagate(), SpliceComponent::Propagate(), SpliceMaxComponent::Propagate(), BlockAffineComponent::Propagate(), SumGroupComponent::Propagate(), PermuteComponent::Propagate(), DctComponent::Propagate(), FixedLinearComponent::Propagate(), FixedAffineComponent::Propagate(), DropoutComponent::Propagate(), and Convolutional1dComponent::Propagate().

                                                                  {
  KALDI_ASSERT((mat.NumRows()  ==  NumRows()) && (mat.NumCols() == NumCols()));
}

ChunkSize()

int32 ChunkSize ( ) const

inline

Definition at line 115 of file nnet-component.h.

References ChunkInfo::num_chunks_, and ChunkInfo::NumRows().

Referenced by SpliceComponent::Backprop(), SpliceMaxComponent::Backprop(), SpliceComponent::Propagate(), and SpliceMaxComponent::Propagate().

{ return NumRows() / num_chunks_; }

GetIndex()

int32 GetIndex

(

int32

offset

)

const

Definition at line 2519 of file nnet-component.cc.

References KALDI_ASSERT.

Referenced by SpliceComponent::Backprop(), SpliceMaxComponent::Backprop(), ChunkInfo::ChunkInfo(), SpliceComponent::Propagate(), and SpliceMaxComponent::Propagate().

                                             {
  if (offsets_.empty()) {  // if data is contiguous
    KALDI_ASSERT((offset <= last_offset_) && (offset >= first_offset_));
    return offset - first_offset_;
  } else  {
    std::vector<int32>::const_iterator iter =
        std::lower_bound(offsets_.begin(), offsets_.end(), offset);
    // make sure offset is present in the vector
    KALDI_ASSERT(iter != offsets_.end() && *iter == offset);
    return static_cast<int32>(iter - offsets_.begin());
  }
}

GetOffset()

int32 GetOffset

(

int32

index

)

const

Definition at line 2532 of file nnet-component.cc.

References KALDI_ASSERT.

Referenced by SpliceComponent::Backprop(), SpliceMaxComponent::Backprop(), ChunkInfo::ChunkInfo(), SpliceComponent::Propagate(), and SpliceMaxComponent::Propagate().

                                            {
  if (offsets_.empty()) { // if data is contiguous
    int32 offset = index + first_offset_;  // just offset by the first_offset_
    KALDI_ASSERT((offset <= last_offset_) && (offset >= first_offset_));
    return offset;
  } else  {
    KALDI_ASSERT((index >= 0) && (index < offsets_.size()));
    return offsets_[index];
  }
}

MakeOffsetsContiguous()

void MakeOffsetsContiguous ( )

inline

Definition at line 111 of file nnet-component.h.

References ChunkInfo::Check(), and ChunkInfo::offsets_.

{ offsets_.clear(); Check(); }

NumChunks()

int32 NumChunks ( ) const

inline

Definition at line 118 of file nnet-component.h.

References ChunkInfo::num_chunks_.

Referenced by SpliceComponent::Backprop(), SpliceMaxComponent::Backprop(), MaxoutComponent::Propagate(), MaxpoolingComponent::Propagate(), PnormComponent::Propagate(), SigmoidComponent::Propagate(), TanhComponent::Propagate(), PowerComponent::Propagate(), SoftHingeComponent::Propagate(), SoftmaxComponent::Propagate(), LogSoftmaxComponent::Propagate(), AffineComponent::Propagate(), SpliceComponent::Propagate(), SpliceMaxComponent::Propagate(), BlockAffineComponent::Propagate(), SumGroupComponent::Propagate(), PermuteComponent::Propagate(), FixedLinearComponent::Propagate(), FixedAffineComponent::Propagate(), DropoutComponent::Propagate(), and Convolutional1dComponent::Propagate().

{ return num_chunks_; }

NumCols()

int32 NumCols ( ) const

inline

Returns the number of columns that we expect the feature matrix to have.

Definition at line 126 of file nnet-component.h.

References ChunkInfo::Check(), ChunkInfo::CheckSize(), and ChunkInfo::feat_dim_.

Referenced by SpliceComponent::Backprop(), SpliceMaxComponent::Backprop(), Component::Propagate(), SpliceComponent::Propagate(), SpliceMaxComponent::Propagate(), DctComponent::Propagate(), and kaldi::nnet2::UnitTestGenericComponentInternal().

{ return feat_dim_; }

NumRows()

int32 NumRows ( ) const

inline

Returns the number of rows that we expect the feature matrix to have.

Definition at line 121 of file nnet-component.h.

References ChunkInfo::first_offset_, ChunkInfo::last_offset_, ChunkInfo::num_chunks_, and ChunkInfo::offsets_.

Referenced by SpliceComponent::Backprop(), SpliceMaxComponent::Backprop(), ChunkInfo::ChunkSize(), Component::Propagate(), DctComponent::Propagate(), and kaldi::nnet2::UnitTestGenericComponentInternal().

                        {
    return num_chunks_ * (!offsets_.empty() ? offsets_.size() :
                                         last_offset_ - first_offset_ + 1); }

Member Data Documentation

feat_dim_

int32 feat_dim_

private

Definition at line 135 of file nnet-component.h.

Referenced by ChunkInfo::NumCols().

first_offset_

int32 first_offset_

private

Definition at line 137 of file nnet-component.h.

Referenced by ChunkInfo::ChunkInfo(), and ChunkInfo::NumRows().

last_offset_

int32 last_offset_

private

Definition at line 140 of file nnet-component.h.

Referenced by ChunkInfo::ChunkInfo(), and ChunkInfo::NumRows().

num_chunks_

int32 num_chunks_

private

Definition at line 136 of file nnet-component.h.

Referenced by ChunkInfo::ChunkSize(), ChunkInfo::NumChunks(), and ChunkInfo::NumRows().

offsets_

std::vector<int32> offsets_

private

Definition at line 141 of file nnet-component.h.

Referenced by ChunkInfo::ChunkInfo(), ChunkInfo::MakeOffsetsContiguous(), and ChunkInfo::NumRows().

---

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

• nnet2/nnet-component.h
• nnet2/nnet-component.cc