|
mxnet
|
A descriptor of a Batch Normalization operation. More...
#include <mkldnn_types.h>
Public Attributes | |
| mkldnn_primitive_kind_t | primitive_kind |
| mkldnn_prop_kind_t | prop_kind |
| mkldnn_memory_desc_t | data_desc |
| Source and destination memory descriptor. More... | |
| mkldnn_memory_desc_t | diff_data_desc |
| Source and destination gradient memory descriptor. More... | |
| mkldnn_memory_desc_t | data_scaleshift_desc |
| mkldnn_memory_desc_t | diff_data_scaleshift_desc |
| mkldnn_memory_desc_t | stat_desc |
| float | batch_norm_epsilon |
| Batch normalization epsilon parameter. More... | |
| unsigned | flags |
A descriptor of a Batch Normalization operation.
| float mkldnn_batch_normalization_desc_t::batch_norm_epsilon |
Batch normalization epsilon parameter.
| mkldnn_memory_desc_t mkldnn_batch_normalization_desc_t::data_desc |
Source and destination memory descriptor.
| mkldnn_memory_desc_t mkldnn_batch_normalization_desc_t::data_scaleshift_desc |
Scale and shift data and gradient memory descriptors.
Scaleshift memory descriptor uses 2D mkldnn_nc format[2,Channels]. 1-st dimension contains gamma parameter, 2-nd dimension contains beta parameter.
| mkldnn_memory_desc_t mkldnn_batch_normalization_desc_t::diff_data_desc |
Source and destination gradient memory descriptor.
| mkldnn_memory_desc_t mkldnn_batch_normalization_desc_t::diff_data_scaleshift_desc |
| unsigned mkldnn_batch_normalization_desc_t::flags |
| mkldnn_primitive_kind_t mkldnn_batch_normalization_desc_t::primitive_kind |
The kind of primitive. Used for self-identifying the primitive descriptor. Must be mkldnn_batch_normalization.
| mkldnn_prop_kind_t mkldnn_batch_normalization_desc_t::prop_kind |
The kind of propagation. Possible values: mkldnn_forward_training, mkldnn_forward_inference, mkldnn_backward, and mkldnn_backward_data.
| mkldnn_memory_desc_t mkldnn_batch_normalization_desc_t::stat_desc |
Statistics memory descriptor.
Statistics (mean or variance) descriptor use 1D mkldnn_x format[Channels].
1.8.11
