dnnl::batch_normalization_forward::desc Struct Reference

Descriptor for a batch normalization forward propagation primitive. More...

#include <dnnl.hpp>

Collaboration diagram for dnnl::batch_normalization_forward::desc:

Public Member Functions
	desc (prop_kind prop_kind, const memory::desc &data_desc, float epsilon, normalization_flags flags)
	Constructs a batch normalization descriptor for forward propagation.

Detailed Description

Descriptor for a batch normalization forward propagation primitive.

Examples

batch_normalization.cpp.

Constructor & Destructor Documentation

desc()

dnnl::batch_normalization_forward::desc::desc ( prop_kind prop_kind, const memory::desc & data_desc, float epsilon, normalization_flags flags )

inline

Constructs a batch normalization descriptor for forward propagation.

Inputs:

• src (dnnl::primitive_desc_base::src_desc(0))
• mean (dnnl::primitive_desc_base::src_desc(1)), if dnnl::normalization_flags::use_global_stats bit-flag is set in flags
• variance (dnnl::primitive_desc_base::src_desc(2)), if dnnl::normalization_flags::use_global_stats bit-flag is set in flags
• scale_and_shift (dnnl::primitive_desc_base::weights_desc(0)), if dnnl::normalization_flags::use_scale_shift bit-flag is set in flags

Outputs:

• dst (dnnl::primitive_desc_base::dst_desc(0))
• mean (dnnl::primitive_desc_base::dst_desc(1)), if dnnl::normalization_flags::use_global_stats bit-flag is not set in flags and prop_kind = dnnl::prop_kind::forward_training
• variance (dnnl::primitive_desc_base::dst_desc(2)), if dnnl::normalization_flags::use_global_stats bit-flag is not set in flags and prop_kind = dnnl::prop_kind::forward_training
• workspace (dnnl::primitive_desc_base::workspace_desc(0)), if dnnl::normalization_flags::fuse_norm_relu bit-flag is set in flags and prop_kind = dnnl::prop_kind::forward_training; must be queried for using primitive_desc_base::query_md() after a corresponding primitive descriptor is created

Note

In-place operation is supported: the dst can refer to the same memory as the src.

Parameters

prop_kind	Propagation kind. Possible values are dnnl::prop_kind::forward_training and dnnl::prop_kind::forward_inference.
data_desc	Source and destination memory descriptors.
epsilon	Batch normalization epsilon parameter.
flags	Batch normalization flags (dnnl::normalization_flags).

---

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

• include/dnnl.hpp