Public Types | Data Fields
InferenceEngine::RNNSequenceLayer Class Reference

Sequence of recurrent cells. More...

#include <ie_layers.h>

Inheritance diagram for InferenceEngine::RNNSequenceLayer:
Inheritance graph
[legend]
Collaboration diagram for InferenceEngine::RNNSequenceLayer:
Collaboration graph
[legend]

Public Types

enum  Direction { FWD, BWD, BDR }
 Direction of iteration through sequence dimension. More...
 
- Public Types inherited from InferenceEngine::RNNCellBase
enum  CellType { LSTM, GRU, RNN, GRU_LBR }
 Direct type of recurrent cell (including subtypes) Description of particular cell semantics is in LSTMCell, GRUCell, RNNCell. More...
 
- Public Types inherited from InferenceEngine::CNNLayer
using Ptr = std::shared_ptr< CNNLayer >
 A shared pointer to CNNLayer.
 

Data Fields

unsigned int axis = 1
 An axis by which iteration is performed. More...
 
Direction direction = FWD
 Direction of iteration through sequence dimension.
 
- Data Fields inherited from InferenceEngine::RNNCellBase
CellType cellType = LSTM
 Direct type of recurrent cell (including subtypes) Description of particular cell semantics is in LSTMCell, GRUCell, RNNCell.
 
int hidden_size = 0
 Size of hidden state data. More...
 
float clip = 0.0f
 Clip data into range [-clip, clip] on input of activations. More...
 
std::vector< std::string > activations
 Activations used inside recurrent cell. More...
 
std::vector< float > activation_alpha
 Alpha parameters of activations. More...
 
std::vector< float > activation_beta
 Beta parameters of activations. More...
 
- Data Fields inherited from InferenceEngine::WeightableLayer
Blob::Ptr _weights
 A pointer to a weights blob.
 
Blob::Ptr _biases
 A pointer to a biases blob.
 
- Data Fields inherited from InferenceEngine::CNNLayer
std::string name
 Layer name.
 
std::string type
 Layer type.
 
Precision precision
 Layer base operating precision.
 
std::vector< DataPtroutData
 A vector of pointers to the output data elements of this layer in the di-graph (order matters)
 
std::vector< DataWeakPtrinsData
 A vector of weak pointers to the input data elements of this layer in the di-graph (order matters)
 
Ptr _fusedWith
 If suggested to fuse - a pointer to the layer which needs to be fused with this layer.
 
UserValue userValue
 Convenience user values to store in this object as extra data.
 
std::string affinity
 Layer affinity set by user.
 
std::map< std::string, std::string > params
 Map of pairs: (parameter name, parameter value)
 
std::map< std::string, Blob::Ptrblobs
 Map of pairs: (name, weights/biases blob)
 

Additional Inherited Members

- Public Member Functions inherited from InferenceEngine::WeightableLayer
 WeightableLayer (const LayerParams &prms)
 A default constructor. Constructs a WeightableLayer instance and initiates layer parameters with the given values. More...
 
- Public Member Functions inherited from InferenceEngine::CNNLayer
 CNNLayer (const LayerParams &prms)
 A constructor. Creates a new CNNLayer instance and initializes layer parameters with the given values. More...
 
virtual ~CNNLayer ()
 A virtual destructor.
 
void fuse (Ptr &layer)
 Sets a layer to be fused with. More...
 
virtual const DataPtr input () const
 Returns the first element of the input data for this layer. More...
 
void validateLayer ()
 Checks if the input data and layer data are legitimate.
 
float GetParamAsFloat (const char *param, float def) const
 Gets float value for the given parameter. More...
 
float GetParamAsFloat (const char *param) const
 Returns a float value for the given layer parameter. More...
 
std::vector< float > GetParamAsFloats (const char *param, std::vector< float > def) const
 Returns a vector of float values for the given parameter or returns the default value. More...
 
std::vector< float > GetParamAsFloats (const char *param) const
 Returns a vector of float values for the given parameter. More...
 
int GetParamAsInt (const char *param, int def) const
 Returns an integer value for the given parameter or returns the default value. More...
 
int GetParamAsInt (const char *param) const
 Returns an integer value for the given parameter. More...
 
std::vector< int > GetParamAsInts (const char *param, std::vector< int > def) const
 Returns a vector of int values for the given parameter or returns the default value. More...
 
std::vector< int > GetParamAsInts (const char *param) const
 Returns a vector of int values for the given parameter. More...
 
unsigned int GetParamAsUInt (const char *param, unsigned int def) const
 Returns an unsigned integer value for the given parameter or returns the default value. More...
 
unsigned int GetParamAsUInt (const char *param) const
 Returns an unsigned integer value for the given parameter. More...
 
std::vector< unsigned int > GetParamAsUInts (const char *param, std::vector< unsigned int > def) const
 Returns a vector of unsigned int values for the given parameter or returns the default value. More...
 
std::vector< unsigned int > GetParamAsUInts (const char *param) const
 Returns a vector of unsigned int values for the given parameter. More...
 
bool GetParamAsBool (const char *param, bool def) const
 Returns a boolean value for the given parameter. More...
 
bool GetParamAsBool (const char *param) const
 Returns a boolean value for the given parameter. More...
 
std::string GetParamAsString (const char *param, const char *def) const
 Returns a string value for the given parameter or returns the default one. More...
 
bool CheckParamPresence (const char *param) const
 Checks the param presence in the layer. More...
 
std::string GetParamAsString (const char *param) const
 Returns a string value for the given parameter. More...
 
std::vector< std::string > GetParamAsStrings (const char *param, std::vector< std::string > def) const
 
- Static Public Member Functions inherited from InferenceEngine::CNNLayer
static float ie_parse_float (const std::string &str)
 Parse string with float in accordance with IE rules. More...
 
static std::string ie_serialize_float (float value)
 serialize float with c_locale formating used for default values serializing
 

Detailed Description

Sequence of recurrent cells.

N - batch size T - sequence size S - state size (=hidden_size) NS - num of state tensors (LSTM=2, GRU/RNN=1) ND - num of direction (BDR=2, WFD/BWD=1)

Inputs: [N,T,D] Xt - input data [ND,N,S] Ht-1 - initial hidden state [ND,N,S] Ct-1 - initial cell state // if NS==2 [N] SL - sequence length (optional input)

Outputs: [ND,N,T,S] Xt - input data [ND,N,S] Ht-1 - initial hidden state [ND,N,S] Ct-1 - initial cell state // if NS==2

NB! if axis==0 batch and sequense dimensions are swapped (N <-> T) for input and output tensors

Weights:

Member Enumeration Documentation

§ Direction

Direction of iteration through sequence dimension.

Enumerator
FWD 

Forward mode. Iterate starts from index 0 with step 1.

BWD 

Backward mode. Iterate starts from last index with step -1.

BDR 

Bidirectional mode. First is forward pass, second is backward.

Field Documentation

§ axis

unsigned int InferenceEngine::RNNSequenceLayer::axis = 1

An axis by which iteration is performed.

axis=0 means first input/output data blob dimension is sequence axis=1 means first input/output data blob dimension is batch


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