.rst .pdf

Contents

enum dnnl::prop_kind

Overview

Propagation kind. More…

#include <dnnl.hpp>

enum prop_kind
{
    undef             = dnnl_prop_kind_undef,
    forward_training  = dnnl_forward_training,
    forward_inference = dnnl_forward_inference,
    forward_scoring   = dnnl_forward_scoring,
    forward           = dnnl_forward,
    backward          = dnnl_backward,
    backward_data     = dnnl_backward_data,
    backward_weights  = dnnl_backward_weights,
    backward_bias     = dnnl_backward_bias,
};

Detailed Documentation

Propagation kind.

Enum Values

undef

Undefined propagation kind.

forward_training

Forward data propagation (training mode).

In this mode, primitives perform computations necessary for subsequent backward propagation.

forward_inference

Forward data propagation (inference mode).

In this mode, primitives perform only computations that are necessary for inference and omit computations that are necessary only for backward propagation.

forward_scoring

Forward data propagation, alias for dnnl::prop_kind::forward_inference.

forward

Forward data propagation, alias for dnnl::prop_kind::forward_training.

backward

Backward propagation (with respect to all parameters).

backward_data

Backward data propagation.

backward_weights

Backward weights propagation.

backward_bias

Backward bias propagation.

enum dnnl::fpmath_mode enum dnnl::scratchpad_mode