IntelĀ® Machine Learning Scaling Library  2018
A library providing an efficient implementation of communication patterns used in deep learning.
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros
Public Member Functions | List of all members
MLSL::Operation Class Reference

A class to hold information about learnable parameters (parameter sets) and activations corresponding to a certain operation of the computational graph. More...

#include <mlsl.hpp>

Public Member Functions

void SetDistribution (Distribution *dist)
 
DistributionGetDistribution ()
 
SessionGetSession ()
 
OpType GetOpType ()
 
void SetPrev (Operation *prev, size_t actIdx, size_t prevOpActIdx)
 
void SetNext (Operation *next, size_t actIdx, size_t nextOpActIdx)
 
const char * GetName ()
 
size_t GetGlobalMinibatchSize ()
 
size_t GetLocalMinibatchSize ()
 
size_t GetGlobalMinibatchOffset ()
 
size_t GetInputCount ()
 
ActivationGetInput (size_t idx)
 
size_t GetOutputCount ()
 
ActivationGetOutput (size_t idx)
 
bool HasParameterSets ()
 
size_t GetParameterSetCount ()
 
ParameterSetGetParameterSet (size_t idx)
 

Detailed Description

A class to hold information about learnable parameters (parameter sets) and activations corresponding to a certain operation of the computational graph.

Member Function Documentation

Distribution* MLSL::Operation::GetDistribution ( )
Returns
The distribution used by the current operation.
size_t MLSL::Operation::GetGlobalMinibatchOffset ( )
Returns
The offset of the local mini-batch portion within the global mini-batch.
size_t MLSL::Operation::GetGlobalMinibatchSize ( )
Returns
The length of the global mini-batch.
Activation* MLSL::Operation::GetInput ( size_t  idx)

Returns the input activation by index.

Parameters
idxthe input activation's index
Returns
The input activation.
size_t MLSL::Operation::GetInputCount ( )
Returns
The number of input activations for the current operation.
size_t MLSL::Operation::GetLocalMinibatchSize ( )
Returns
The length of the local mini-batch portion.
const char* MLSL::Operation::GetName ( )
Returns
The operation name.
OpType MLSL::Operation::GetOpType ( )
Returns
The operation type.
Activation* MLSL::Operation::GetOutput ( size_t  idx)

Returns the output activation by index.

Parameters
idxthe output activation's index
Returns
The output activation.
size_t MLSL::Operation::GetOutputCount ( )
Returns
The number of output activations for the current operation.
ParameterSet* MLSL::Operation::GetParameterSet ( size_t  idx)

Returns the parameter set by index

Parameters
idxthe parameter set's index
Returns
The parameter set.
size_t MLSL::Operation::GetParameterSetCount ( )
Returns
The number of parameter sets for the current operation.
Session* MLSL::Operation::GetSession ( )
Returns
The session to which the current operation belongs.
bool MLSL::Operation::HasParameterSets ( )
Returns
True if the current operation has parameter sets (weights or bias), false otherwise.
void MLSL::Operation::SetDistribution ( Distribution dist)

Sets the operation's distribution for the case when the operation has been created without the distribution through session->AddOperation(regInfo, NULL)

void MLSL::Operation::SetNext ( Operation next,
size_t  actIdx,
size_t  nextOpActIdx 
)

Sets the next operation in the computational graph.

Parameters
nextthe next operation
actIdxthe index of the current operation's output activation correspoding to the next operation
nextOpActIdxthe index of the next operation's input activation correspoding to the current operation
void MLSL::Operation::SetPrev ( Operation prev,
size_t  actIdx,
size_t  prevOpActIdx 
)

Sets the previous operation in the computational graph.

Parameters
prevthe previous operation
actIdxthe index of the current operation's input activation correspoding to the previous operation
prevOpActIdxthe index of the previous operation's output activation correspoding to the current operation

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