A descriptor of a element-wise operation. More...
#include <dnnl_types.h>
Public Attributes | |
dnnl_primitive_kind_t | primitive_kind |
The kind of primitive. More... | |
dnnl_prop_kind_t | prop_kind |
The kind of propagation. More... | |
dnnl_alg_kind_t | alg_kind |
The kind of eltwise algorithm. More... | |
dnnl_memory_desc_t | data_desc |
Source and destination memory descriptor. | |
dnnl_memory_desc_t | diff_data_desc |
Source and destination gradient memory descriptor. | |
float | alpha |
Algorithm specific parameter. More... | |
A descriptor of a element-wise operation.
dnnl_primitive_kind_t dnnl_eltwise_desc_t::primitive_kind |
The kind of primitive.
Used for self-identifying the primitive descriptor. Must be dnnl_eltwise.
dnnl_prop_kind_t dnnl_eltwise_desc_t::prop_kind |
The kind of propagation.
Possible values: dnnl_forward_training, dnnl_forward_inference, dnnl_backward, and dnnl_backward_data.
dnnl_alg_kind_t dnnl_eltwise_desc_t::alg_kind |
The kind of eltwise algorithm.
Possible values: dnnl_eltwise_relu, dnnl_eltwise_tanh, dnnl_eltwise_elu, dnnl_eltwise_square, dnnl_eltwise_abs, dnnl_eltwise_sqrt, dnnl_eltwise_linear, dnnl_eltwise_bounded_relu, dnnl_eltwise_soft_relu, dnnl_eltwise_logistic, dnnl_eltwise_exp, dnnl_eltwise_gelu_tanh, dnnl_eltwise_swish, dnnl_eltwise_log, dnnl_eltwise_clip, dnnl_eltwise_pow, dnnl_eltwise_gelu_erf. Possible values for passing destination memory on backward: dnnl_eltwise_relu_use_dst_for_bwd, dnnl_eltwise_tanh_use_dst_for_bwd, dnnl_eltwise_elu_use_dst_for_bwd, dnnl_eltwise_sqrt_use_dst_for_bwd, dnnl_eltwise_logistic_use_dst_for_bwd, dnnl_eltwise_exp_use_dst_for_bwd.
float dnnl_eltwise_desc_t::alpha |
Algorithm specific parameter.
Accordance table:
alpha
– negative slope, beta
ignoredalpha
and beta
ignoredalpha
– negative slope, beta
ignoredalpha
and beta
ignoredalpha
and beta
ignoredalpha
and beta
ignoredalpha
– scale, beta
– shiftalpha
– upper bound, beta
ignoredalpha
and beta
ignoredalpha
and beta
ignoredalpha
and beta
ignoredalpha
and beta
ignoredalpha
– sigmoid arg scaling, beta
ignoredalpha
and beta
ignoredalpha
– lower bound, beta
– upper boundalpha
– scale, beta
– exponentalpha
and beta
ignored