Package | Description |
---|---|
org.bytedeco.mkldnn | |
org.bytedeco.mkldnn.global |
Modifier and Type | Method and Description |
---|---|
batch_normalization_forward.desc |
batch_normalization_forward.desc.data(mkldnn_batch_normalization_desc_t setter) |
batch_normalization_backward.desc |
batch_normalization_backward.desc.data(mkldnn_batch_normalization_desc_t setter) |
Modifier and Type | Method and Description |
---|---|
static int |
mkldnn.mkldnn_batch_normalization_backward_desc_init(mkldnn_batch_normalization_desc_t bnrm_desc,
int prop_kind,
mkldnn_memory_desc_t diff_data_desc,
mkldnn_memory_desc_t data_desc,
float epsilon,
int flags)
Initializes a batch normalization descriptor \p bnrm_desc for backward
propagation with respect to data and scale-shift parameters using memory
descriptors \p data_desc and \p diff_data_desc, normalization parameter
\p epsilon, and \p flags set using bit flags of type
mkldnn_batch_normalization_desc_t.
|
static int |
mkldnn.mkldnn_batch_normalization_forward_desc_init(mkldnn_batch_normalization_desc_t bnrm_desc,
int prop_kind,
mkldnn_memory_desc_t data_desc,
float epsilon,
int flags)
Initializes a batch normalization descriptor \p bnrm_desc for forward
propagation using \p prop_kind (possible values are
#mkldnn_forward_training and #mkldnn_forward_inference), memory descriptor
\p data_desc, normalization parameter \p epsilon, and \p flags set using bit
flags of type mkldnn_batch_normalization_desc_t.
|
Copyright © 2022. All rights reserved.