DML_ACTIVATION_RELU_GRAD_OPERATOR_DESC structure (directml.h)

Computes backpropagation gradients for a rectified linear unit (ReLU). This operator performs the following element-wise computation.

X = InputTensor
dY = InputGradientTensor

OutputGradientTensor = (X > 0 ? dY : 0)

The corresponding forward-pass operator is DML_ACTIVATION_RELU_OPERATOR_DESC.

Syntax

struct DML_ACTIVATION_RELU_GRAD_OPERATOR_DESC {
  const DML_TENSOR_DESC *InputTensor;
  const DML_TENSOR_DESC *InputGradientTensor;
  const DML_TENSOR_DESC *OutputGradientTensor;
};

Members

InputTensor

Type: const DML_TENSOR_DESC*

The input (feature) tensor. This is typically the same input as was provided during the forward pass (see DML_ACTIVATION_RELU_OPERATOR_DESC).

InputGradientTensor

Type: const DML_TENSOR_DESC*

The incoming gradient tensor. This is typically obtained from the output of backpropagation of a preceding layer. The Sizes and DataType of this tensor must exactly match those of the InputTensor.

OutputGradientTensor

Requirements

Requirement	Value
Minimum supported client	Windows 10 Build 20348
Minimum supported server	Windows 10 Build 20348
Header	directml.h

See also

DML_ACTIVATION_RELU_OPERATOR_DESC