Intel(R) Math Kernel Library for Deep Neural Networks (Intel(R) MKL-DNN)
0.21.0
Performance library for Deep Learning
|
#include <mkldnn_types.h>
Public Attributes | |
mkldnn_alg_kind_t | cell_kind |
RNN cell kind. More... | |
mkldnn_alg_kind_t | activation_kind |
Activation function used. More... | |
unsigned int | flags |
RNN cell flags. More... | |
float | alpha |
alpha is a negative slope parameter (used only if (flags & mkldnn_rnn_cell_with_relu) != 0 ) More... | |
float | clipping |
clipping parameter (used only if (flags & mkldnn_rnn_cell_with_clipping) != 0 ) More... | |
mkldnn_alg_kind_t mkldnn_rnn_cell_desc_t::cell_kind |
RNN cell kind.
Must be one of mkldnn_vanilla_rnn, mkldnn_vanilla_lstm, mkldnn_vanilla_gru, or mkldnn_gru_linear_before_reset.
mkldnn_alg_kind_t mkldnn_rnn_cell_desc_t::activation_kind |
Activation function used.
Must be either mkldnn_eltwise_relu or mkldnn_eltwise_tanh.
unsigned int mkldnn_rnn_cell_desc_t::flags |
RNN cell flags.
float mkldnn_rnn_cell_desc_t::alpha |
alpha
is a negative slope parameter (used only if (flags & mkldnn_rnn_cell_with_relu) != 0
)
float mkldnn_rnn_cell_desc_t::clipping |
clipping parameter (used only if (flags & mkldnn_rnn_cell_with_clipping) != 0
)