Deep Neural Network Library (DNNL)  1.1.3
Performance library for Deep Learning
Public Member Functions | List of all members
dnnl::lstm_backward::primitive_desc Struct Reference

Primitive descriptor for LSTM backward propagation. More...

#include <dnnl.hpp>

Inherits dnnl::rnn_primitive_desc_base.

Public Member Functions

 primitive_desc (dnnl_primitive_desc_t pd)
 Initializes a primitive descriptor for LSTM backward propagation from a C primitive descriptor pd. More...
 
memory::desc src_layer_desc () const
 Queries source layer memory descriptor.
 
memory::desc src_iter_desc () const
 Queries source recurrent hidden state memory descriptor. More...
 
memory::desc src_iter_c_desc () const
 Queries source recurrent cell state memory descriptor.
 
memory::desc weights_layer_desc () const
 Queries weights layer memory descriptor.
 
memory::desc weights_iter_desc () const
 Queries weights iteration memory descriptor.
 
memory::desc bias_desc () const
 Queries bias memory descriptor. More...
 
memory::desc dst_layer_desc () const
 Queries destination layer memory descriptor.
 
memory::desc dst_iter_desc () const
 Queries destination recurrent hidden state memory descriptor. More...
 
memory::desc dst_iter_c_desc () const
 Queries destination recurrent cell state memory descriptor.
 
memory::desc workspace_desc () const
 Queries workspace memory descriptor. More...
 
memory::desc diff_src_layer_desc () const
 Queries diff source layer memory descriptor.
 
memory::desc diff_src_iter_desc () const
 Queries diff source recurrent hidden state memory descriptor. More...
 
memory::desc diff_src_iter_c_desc () const
 Queries diff source recurrent cell state memory descriptor.
 
memory::desc diff_weights_layer_desc () const
 Queries diff weights layer memory descriptor.
 
memory::desc diff_weights_iter_desc () const
 Queries diff weights iteration memory descriptor.
 
memory::desc diff_bias_desc () const
 Queries diff bias memory descriptor.
 
memory::desc diff_dst_layer_desc () const
 Queries diff destination layer memory descriptor.
 
memory::desc diff_dst_iter_desc () const
 Queries diff destination recurrent hidden state memory descriptor. More...
 
memory::desc diff_dst_iter_c_desc () const
 Queries diff destination recurrent cell state memory descriptor.
 

Detailed Description

Primitive descriptor for LSTM backward propagation.

Examples:
rnn_training_f32.cpp.

Constructor & Destructor Documentation

◆ primitive_desc()

dnnl::lstm_backward::primitive_desc::primitive_desc ( dnnl_primitive_desc_t  pd)
inline

Initializes a primitive descriptor for LSTM backward propagation from a C primitive descriptor pd.

Member Function Documentation

◆ src_iter_desc()

memory::desc dnnl::lstm_backward::primitive_desc::src_iter_desc ( ) const
inline

Queries source recurrent hidden state memory descriptor.

Returns a zero_md if no src_iter was specified at op_desc creation time.

◆ bias_desc()

memory::desc dnnl::lstm_backward::primitive_desc::bias_desc ( ) const
inline

Queries bias memory descriptor.

Returns a zero_md if no bias was specified at op_desc creation time.

◆ dst_iter_desc()

memory::desc dnnl::lstm_backward::primitive_desc::dst_iter_desc ( ) const
inline

Queries destination recurrent hidden state memory descriptor.

Returns a zero_md if no dst_iter was specified at op_desc creation time.

◆ workspace_desc()

memory::desc dnnl::lstm_backward::primitive_desc::workspace_desc ( ) const
inline

Queries workspace memory descriptor.

Returns a zero_md if no worspace is required.

◆ diff_src_iter_desc()

memory::desc dnnl::lstm_backward::primitive_desc::diff_src_iter_desc ( ) const
inline

Queries diff source recurrent hidden state memory descriptor.

Returns a zero_md if no diff_src_iter was specified at op_desc creation time.

◆ diff_dst_iter_desc()

memory::desc dnnl::lstm_backward::primitive_desc::diff_dst_iter_desc ( ) const
inline

Queries diff destination recurrent hidden state memory descriptor.

Returns a zero_md if no diff_dst_iter was specified at op_desc creation time.


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