Dragon - C++ API
A Computation Graph Virtual Machine Based Deep Learning Framework
Public Member Functions | Public Attributes | Protected Attributes | List of all members
dragon::CuDNNDepthwiseConv2dOp< Context > Class Template Referencefinal

#include <depthwise_conv_op.h>

Inheritance diagram for dragon::CuDNNDepthwiseConv2dOp< Context >:
dragon::DepthwiseConv2dOp< Context > dragon::ConvOpBase< Context > dragon::Operator< Context > dragon::OperatorBase

Public Member Functions

 CuDNNDepthwiseConv2dOp (const OperatorDef &def, Workspace *ws)
 ~CuDNNDepthwiseConv2dOp ()
void RunOnDevice () override
 Implement the detailed execution. More...
template<typename T >
void RunImpl ()
- Public Member Functions inherited from dragon::DepthwiseConv2dOp< Context >
 DepthwiseConv2dOp (const OperatorDef &def, Workspace *ws)
bool Transposed () override
bool HasBias () override
template<typename T >
void RunImpl ()
- Public Member Functions inherited from dragon::ConvOpBase< Context >
 ConvOpBase (const OperatorDef &def, Workspace *ws)
 DECLARE_ARGS_WITH_DESC (int64_t, output_shape)
 DECLARE_ARGS_WITH_DESC (int64_t, output_padding)
void Setup (int num_axes)
void Reshape (bool backward=false)
template<typename T >
void Wx (const T *, const T *, T *, bool)
template<typename T >
void Pb (const T *, T *)
template<typename T >
void Dx (const T *, const T *, T *)
template<typename T >
void Dw (const T *, const T *, T *, bool)
template<typename T >
void Db (const T *, T *)
- Public Member Functions inherited from dragon::Operator< Context >
 Operator (const OperatorDef &def, Workspace *ws)
 Default constructor. More...
void Run (int stream_id=0) final
 Run this operator on the specified stream. More...
virtual void PrepareResource ()
 Prepare the content of inputs. More...
virtual void ReleaseResource ()
 Release the ownership of inputs. More...
virtual void MemorySwitch ()
 Coordinate the context of inputs and outputs. More...
Context * ctx ()
 Return the internal context. More...
bool AllowRun ()
 Whether this operator can be ignored. More...
- Public Member Functions inherited from dragon::OperatorBase
 OperatorBase (const OperatorDef &def, Workspace *ws)
 Default constructor. More...
virtual ~OperatorBase ()
 Default deconstructor. More...
TensorX (int i)
 Return the specified input tensor. More...
TensorY (int i)
 Return the specified output tensor. More...
int XSize ()
 Return the number of inputs. More...
int YSize ()
 Return the number of outputs. More...
void UpdateFrom (const OperatorDef &def)
 Modify operator according to the given def. More...
void SwitchToPhase (const string &phase)
 Switch the internal running phase. More...
virtual void Fusion (void *graph)
 Fusion this operator into the specified graph. More...
const string & name () const
 Return the operator name. More...
const string & type () const
 Return the operator type. More...
const string & phase () const
 Return the current running phase. More...
const string & handle () const
 Return the resource handle. More...
const string & dtype () const
 Return the data type. More...
const string & data_format () const
 Return the data format. More...
const string unique_name (const string &name) const
 Return the unique name in this operator. More...
Workspacews () const
 Return the parent workspace. More...
template<typename T >
Arg (const string &name, const T &default_value)
 Return the value of the specified argument. More...
template<typename T >
vector< T > Args (const string &name)
 Return the values of the specified argument. More...
const Map< std::string, const Argument * > & args ()
 Return the argument map. More...
const Argument & arg (const string &name)
 Return the specified argument. More...
SubGraphsubgraph ()
 Return the recomputing subgraph. More...
void set_subgraph (SubGraph subgraph)
 Set the given recomputing subgraph. More...
const OperatorDef & def () const
 Return the stored def. More...
string DebugString () const
 Return the debug string of stored def. More...
string DTypeString (const Tensor &, const Set< string > &) const
 Return the dtype string according to given tensor. More...
string DTypeString (const string &, const Set< string > &) const

Public Attributes

- Public Attributes inherited from dragon::DepthwiseConv2dOp< Context >
- Public Attributes inherited from dragon::ConvOpBase< Context >
vec64_t kshape_
vec64_t stride_
vec64_t pad_l_
vec64_t pad_r_
vec64_t dilation_
vec64_t in_shape_
vec64_t out_shape_
vec64_t x_shape_
vec64_t y_shape_
vec64_t w_shape_
vec64_t b_shape_
string padding_
int64_t is_1x1_
int64_t num_output_
int64_t group_
int64_t axis_
int64_t num_axes_
int64_t channels_
int64_t out_dim_
int64_t conv_in_channels_
int64_t conv_out_channels_
int64_t conv_out_dim_
int64_t kernel_dim_
int64_t col_dim_
int64_t col_ofs_
int64_t output_ofs_
int64_t w_ofs_
int64_t x_ofs_
int64_t y_ofs_

Protected Attributes

cudnnTensorDescriptor_t bias_desc_
cudnnTensorDescriptor_t output_desc_
- Protected Attributes inherited from dragon::Operator< Context >
Context ctx_
 Store the internal context. More...
bool allow_run_
bool allow_recomp_
bool do_sync_
- Protected Attributes inherited from dragon::OperatorBase
 Store the parent workspace. More...
OperatorDef def_
 Store the def. More...
SubGraph subgraph_
 Store the recomputing subgraph. More...
string phase_
 Store the phase and handle. More...
string handle_
string dtype_
 Store the data type and format. More...
string data_format_
vector< Tensor * > inputs_
 Store the pointer of inputs and outputs. More...
vector< Tensor * > outputs_
Map< string, const Argument * > args_
 Store the defined arguments. More...

Additional Inherited Members

- Public Types inherited from dragon::OperatorBase
typedef Map< string, vector< OperatorBase * > > SubGraph

Constructor & Destructor Documentation

◆ CuDNNDepthwiseConv2dOp()

template<class Context >
dragon::CuDNNDepthwiseConv2dOp< Context >::CuDNNDepthwiseConv2dOp ( const OperatorDef &  def,
Workspace ws 

◆ ~CuDNNDepthwiseConv2dOp()

template<class Context >
dragon::CuDNNDepthwiseConv2dOp< Context >::~CuDNNDepthwiseConv2dOp ( )

Member Function Documentation

◆ RunImpl()

template<class Context >
template<typename T >
void dragon::CuDNNDepthwiseConv2dOp< Context >::RunImpl ( )

◆ RunOnDevice()

template<class Context >
void dragon::CuDNNDepthwiseConv2dOp< Context >::RunOnDevice ( )

Implement the detailed execution.

Reimplemented from dragon::DepthwiseConv2dOp< Context >.

Member Data Documentation

◆ bias_desc_

template<class Context >
cudnnTensorDescriptor_t dragon::CuDNNDepthwiseConv2dOp< Context >::bias_desc_

◆ output_desc_

template<class Context >
cudnnTensorDescriptor_t dragon::CuDNNDepthwiseConv2dOp< Context >::output_desc_


template<class Context >
dragon::CuDNNDepthwiseConv2dOp< Context >::USE_CONVOLUTION_FUNCTIONS


template<class Context >
dragon::CuDNNDepthwiseConv2dOp< Context >::USE_OPERATOR_FUNCTIONS

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