Deep Neural Network Library (DNNL)  1.3.0
Performance library for Deep Learning
Public Member Functions | List of all members
dnnl::layer_normalization_backward::desc Struct Reference

Descriptor for a layer normalization backward propagation primitive. More...

#include <dnnl.hpp>

Collaboration diagram for dnnl::layer_normalization_backward::desc:
Collaboration graph
[legend]

Public Member Functions

 desc (prop_kind prop_kind, const memory::desc &diff_data_desc, const memory::desc &data_desc, const memory::desc &stat_desc, float epsilon, normalization_flags flags)
 Constructs a descriptor for layer normalization backward propagation primitive. More...
 
 desc (prop_kind prop_kind, const memory::desc &diff_data_desc, const memory::desc &data_desc, float epsilon, normalization_flags flags)
 Constructs a descriptor for layer normalization backward propagation primitive. More...
 

Detailed Description

Descriptor for a layer normalization backward propagation primitive.

Constructor & Destructor Documentation

◆ desc() [1/2]

dnnl::layer_normalization_backward::desc::desc ( prop_kind  prop_kind,
const memory::desc diff_data_desc,
const memory::desc data_desc,
const memory::desc stat_desc,
float  epsilon,
normalization_flags  flags 
)
inline

Constructs a descriptor for layer normalization backward propagation primitive.

Inputs:

Outputs:

Parameters
prop_kindPropagation kind. Possible values are dnnl::prop_kind::backward_data and dnnl::prop_kind::backward (diffs for all parameters are computed in this case).
diff_data_descDiff source and diff destination memory descriptor.
data_descSource memory descriptor.
stat_descStatistics memory descriptors.
epsilonLayer normalization epsilon parameter.
flagsLayer normalization flags (dnnl::normalization_flags).

◆ desc() [2/2]

dnnl::layer_normalization_backward::desc::desc ( prop_kind  prop_kind,
const memory::desc diff_data_desc,
const memory::desc data_desc,
float  epsilon,
normalization_flags  flags 
)
inline

Constructs a descriptor for layer normalization backward propagation primitive.

Inputs:

Outputs:

Parameters
prop_kindPropagation kind. Possible values are dnnl::prop_kind::backward_data and dnnl::prop_kind::backward (diffs for all parameters are computed in this case).
diff_data_descDiff source and diff destination memory descriptor.
data_descSource memory descriptor.
epsilonLayer normalization epsilon parameter.
flagsLayer normalization flags (dnnl::normalization_flags).

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