ONE - On-device Neural Engine
Loading...
Searching...
No Matches
onert::backend::train::ITensorRegistry Class Referenceabstract

#include <ITensorRegistry.h>

Collaboration diagram for onert::backend::train::ITensorRegistry:

Public Member Functions

virtual ITensorgetBackPropITensor (const ir::OperandIndex &)=0
 Returns pointer of ITensor for back propatation.
 
virtual ITensorgetGradientITensor (const ir::OperandIndex &)=0
 Returns pointer of ITensor for gradient.
 
virtual void iterateTrainableTensors (const std::function< void(const ir::OperandIndex &, const train::ITrainableTensor *)> &) const =0
 Iterate ITrainableTensors with fn.
 
- Public Member Functions inherited from onert::backend::ITensorRegistry
virtual ~ITensorRegistry ()=default
 Deconstruct itself.
 
virtual ITensorgetITensor (const ir::OperandIndex &)=0
 Returns pointer of ITensor among native and migrant tensors.
 
virtual ITensorgetNativeITensor (const ir::OperandIndex &)=0
 Returns pointer of ITensor among native tensors.
 
virtual bool setMigrantTensor (const ir::OperandIndex &, IPortableTensor *)
 Set the Migrant Tensor which are from other backends.
 

Detailed Description

Definition at line 30 of file ITensorRegistry.h.

Member Function Documentation

◆ getBackPropITensor()

virtual ITensor * onert::backend::train::ITensorRegistry::getBackPropITensor ( const ir::OperandIndex )
pure virtual

Returns pointer of ITensor for back propatation.

Note
Return tensor cannot be used longer than dynamic tensor manager

Implemented in onert::backend::train::PortableTensorRegistryTemplate< Tensor, TrainableTensor, BackPropTensor, GradientTensor >, and onert::backend::builtin::train::TensorRegistry.

◆ getGradientITensor()

virtual ITensor * onert::backend::train::ITensorRegistry::getGradientITensor ( const ir::OperandIndex )
pure virtual

Returns pointer of ITensor for gradient.

Note
Returned tensor cannot be used longer than dynamic tensor manager

Implemented in onert::backend::train::PortableTensorRegistryTemplate< Tensor, TrainableTensor, BackPropTensor, GradientTensor >, and onert::backend::builtin::train::TensorRegistry.

◆ iterateTrainableTensors()

virtual void onert::backend::train::ITensorRegistry::iterateTrainableTensors ( const std::function< void(const ir::OperandIndex &, const train::ITrainableTensor *)> &  ) const
pure virtual

Iterate ITrainableTensors with fn.

Parameters
fnfunction to be called with OperandIndex and a pointer to ITrainableTensor

Implemented in onert::backend::train::PortableTensorRegistryTemplate< Tensor, TrainableTensor, BackPropTensor, GradientTensor >.


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