ONE - On-device Neural Engine
|
Class to gather NN package's executor set. More...
#include <IExecutors.h>
Public Member Functions | |
virtual | ~IExecutors ()=default |
Virtual IExecutors destructor. | |
virtual void | emplace (const ir::ModelIndex &model_index, const ir::SubgraphIndex &subg_index, std::unique_ptr< IExecutor > exec)=0 |
Insert executor in executor set. | |
virtual IExecutor * | at (const ir::ModelIndex &model_index, const ir::SubgraphIndex &subg_index) const =0 |
Return executor of index. | |
IExecutor * | entryExecutor () const |
virtual uint32_t | inputSize () const =0 |
Return executor set's number of input. | |
virtual uint32_t | outputSize () const =0 |
Return executor set's number of output. | |
virtual const ir::OperandInfo & | inputInfo (const ir::IOIndex &index) const =0 |
Return NN package input tensor info. | |
virtual const ir::OperandInfo & | outputInfo (const ir::IOIndex &index) const =0 |
Return NN package output tensor info. | |
virtual void | execute (const ExecutionContext &ctx)=0 |
Execute NN package executor set. | |
Class to gather NN package's executor set.
Definition at line 30 of file IExecutors.h.
|
virtualdefault |
Virtual IExecutors destructor.
|
pure virtual |
Return executor of index.
[in] | model_index | Model index |
[in] | subg_index | Subgraph index |
Implemented in onert::exec::MultiModelExecutors, onert::exec::SingleModelExecutors, and onert::exec::train::TrainableExecutors.
Referenced by entryExecutor(), onert::backend::builtin::kernel::IfLayer::run(), and onert::backend::builtin::kernel::WhileLayer::run().
|
pure virtual |
Insert executor in executor set.
[in] | model_index | Model index |
[in] | subg_index | Subgraph index |
[in] | exec | Executor to insert |
Implemented in onert::exec::MultiModelExecutors, onert::exec::SingleModelExecutors, and onert::exec::train::TrainableExecutors.
|
inline |
Definition at line 60 of file IExecutors.h.
References at().
Referenced by onert::exec::SingleModelExecutors::execute(), onert::exec::SingleModelExecutors::inputInfo(), onert::exec::SingleModelExecutors::inputSize(), onert::exec::SingleModelExecutors::outputInfo(), onert::exec::SingleModelExecutors::outputSize(), onert::backend::builtin::kernel::IfLayer::run(), and onert::backend::builtin::kernel::WhileLayer::run().
|
pure virtual |
Execute NN package executor set.
[in] | ctx | Execution context |
Implemented in onert::exec::MultiModelExecutors, onert::exec::SingleModelExecutors, and onert::exec::train::TrainableExecutors.
|
pure virtual |
Return NN package input tensor info.
[in] | index | Input index |
Implemented in onert::exec::MultiModelExecutors, onert::exec::SingleModelExecutors, and onert::exec::train::TrainableExecutors.
|
pure virtual |
Return executor set's number of input.
Implemented in onert::exec::MultiModelExecutors, onert::exec::SingleModelExecutors, and onert::exec::train::TrainableExecutors.
|
pure virtual |
Return NN package output tensor info.
[in] | index | Output index |
Implemented in onert::exec::MultiModelExecutors, onert::exec::SingleModelExecutors, and onert::exec::train::TrainableExecutors.
|
pure virtual |
Return executor set's number of output.
Implemented in onert::exec::MultiModelExecutors, onert::exec::SingleModelExecutors, and onert::exec::train::TrainableExecutors.