Deep Neural Network Library (DNNL)  1.1.3
Performance library for Deep Learning
Public Member Functions | List of all members
dnnl::gru_forward::desc Struct Reference

Descriptor for GRU forward propagation. More...

#include <dnnl.hpp>

Collaboration diagram for dnnl::gru_forward::desc:
Collaboration graph
[legend]

Public Member Functions

 desc (prop_kind aprop_kind, rnn_direction direction, const memory::desc &src_layer_desc, const memory::desc &src_iter_desc, const memory::desc &weights_layer_desc, const memory::desc &weights_iter_desc, const memory::desc &bias_desc, const memory::desc &dst_layer_desc, const memory::desc &dst_iter_desc, rnn_flags flags=rnn_flags::undef)
 Initializes a GRU descriptor for forward propagation using prop_kind, direction, and memory descriptors. More...
 

Detailed Description

Descriptor for GRU forward propagation.

Constructor & Destructor Documentation

◆ desc()

dnnl::gru_forward::desc::desc ( prop_kind  aprop_kind,
rnn_direction  direction,
const memory::desc src_layer_desc,
const memory::desc src_iter_desc,
const memory::desc weights_layer_desc,
const memory::desc weights_iter_desc,
const memory::desc bias_desc,
const memory::desc dst_layer_desc,
const memory::desc dst_iter_desc,
rnn_flags  flags = rnn_flags::undef 
)
inline

Initializes a GRU descriptor for forward propagation using prop_kind, direction, and memory descriptors.

Note
If prop_kind equals dnnl::forward_training, you must query a workspace memory descriptor before creating the primitive.

flags is a parameter to the GRU descriptor and is currently ignored.

src_iter_desc, bias_desc, and dst_iter_desc are allowed to point to a zero memory descriptor, which would indicate that the GRU primitive should not use them and will default to zero values.

Note
All memory descriptors except src_iter_desc can be initialized with an dnnl::memory::format_tag::any value of format_kind.

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