Caffe2 - C++ API
A deep learning, cross platform ML framework
Public Member Functions | Data Fields
caffe2::AliasOp< Context > Class Template Referencefinal

Alias op makes the output and the input share the same underlying storage. More...

#include <utility_ops.h>

Inheritance diagram for caffe2::AliasOp< Context >:
caffe2::Operator< Context > caffe2::OperatorBase caffe2::Observable< OperatorBase >

Public Member Functions

 USE_SIMPLE_CTOR_DTOR (AliasOp)
 
bool RunOnDevice () override
 
- Public Member Functions inherited from caffe2::Operator< Context >
 Operator (const OperatorDef &operator_def, Workspace *ws)
 
const Tensor< Context > & Input (int idx)
 
Tensor< Context > * Output (int idx)
 
void WaitEvent (const Event &ev, int stream_id=-1) final
 
void WaitEvents (const std::vector< const Event * > &events, int stream_id=-1) final
 
bool Run (int stream_id=0) final
 
bool RunAsync (int stream_id=0) final
 
bool IsStreamFree (int stream_id) const override
 
bool HasAsyncPart () const override
 
bool SupportsAsyncScheduling () const override
 
const Context * getContext () const
 
- Public Member Functions inherited from caffe2::OperatorBase
 OperatorBase (const OperatorDef &operator_def, Workspace *ws)
 
bool HasArgument (const string &name) const
 Checks if the operator has an argument of the given name.
 
template<typename T >
GetSingleArgument (const string &name, const T &default_value) const
 
template<typename T >
bool HasSingleArgumentOfType (const string &name) const
 
template<typename T >
vector< T > GetRepeatedArgument (const string &name, const vector< T > &default_value={}) const
 
template<typename T >
const T & Input (int idx)
 
template<typename T >
T * Output (int idx)
 
template<typename T >
T * Output (int idx, T *allocated)
 
const BlobInputBlob (int idx)
 
BlobOutputBlob (int idx)
 
template<typename T >
bool InputIsType (int idx)
 
template<typename T >
bool OutputIsType (int idx)
 
int InputSize () const
 
int OutputSize () const
 
const vector< const Blob * > & Inputs () const
 
const vector< Blob * > & Outputs ()
 
vector< TensorShape > InputTensorShapes ()
 
void Wait (const OperatorBase &other, int stream_id=-1)
 
virtual void Finish ()
 
virtual void AddRelatedBlobInfo (EnforceNotMet *err)
 
const OperatorDef & debug_def () const
 
void set_debug_def (const std::shared_ptr< const OperatorDef > &operator_def)
 
bool has_debug_def () const
 
void RecordLastFailedOpNetPosition ()
 
int net_position () const
 
void set_net_position (int idx)
 
const DeviceOption & device_option () const
 
const Eventevent () const
 
Eventevent ()
 
void ResetEvent ()
 
void DisableEvent ()
 
bool IsEventDisabled () const
 
const std::string & type () const
 
void annotate_engine (const std::string &engine)
 
const std::string & engine () const
 
- Public Member Functions inherited from caffe2::Observable< OperatorBase >
const ObserverAttachObserver (std::unique_ptr< Observer > observer)
 
std::unique_ptr< ObserverDetachObserver (const Observer *observer_ptr)
 Returns a unique_ptr to the removed observer. More...
 
virtual size_t NumObservers ()
 
void StartAllObservers ()
 
void StopAllObservers ()
 

Data Fields

 USE_OPERATOR_CONTEXT_FUNCTIONS
 

Additional Inherited Members

- Public Types inherited from caffe2::Observable< OperatorBase >
using Observer = ObserverBase< OperatorBase >
 
- Static Public Attributes inherited from caffe2::OperatorBase
static constexpr int kNoNetPositionSet = -1
 
- Protected Member Functions inherited from caffe2::Operator< Context >
void RecordEvent (const char *err_msg=nullptr) final
 
std::string getErrorMsg ()
 
- Protected Member Functions inherited from caffe2::OperatorBase
 DISABLE_COPY_AND_ASSIGN (OperatorBase)
 
- Protected Attributes inherited from caffe2::Operator< Context >
Context context_
 
- Protected Attributes inherited from caffe2::OperatorBase
std::unique_ptr< Eventevent_
 
- Protected Attributes inherited from caffe2::Observable< OperatorBase >
std::vector< std::unique_ptr< Observer > > observers_list_
 

Detailed Description

template<class Context>
class caffe2::AliasOp< Context >

Alias op makes the output and the input share the same underlying storage.

WARNING: in general, in caffe2's operator interface different tensors should have different underlying storage, which is the assumption made by components such as the dependency engine and memory optimization. Thus, in normal situations you should not use the AliasOp, especially in a normal forward-backward pass.

The Alias op is provided so one can achieve true asynchrony, such as Hogwild, in a graph. But make sure you understand all the implications similar to multi-thread computation before you use it explicitly.

Definition at line 164 of file utility_ops.h.


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