Define Abs_Ex op in NeuralNetworksEx.h (#3758)
authorShubham Gupta/SNAP /SRI-Bangalore/Engineer/삼성전자 <shub98.gupta@samsung.com>
Fri, 30 Nov 2018 07:38:08 +0000 (13:08 +0530)
committer오형석/동작제어Lab(SR)/Staff Engineer/삼성전자 <hseok82.oh@samsung.com>
Fri, 30 Nov 2018 07:38:08 +0000 (16:38 +0900)
This patch will define Abs_Ex op in NeuralNetworksEx.h

Signed-off-by: shubham <shub98.gupta@samsung.com>
include/NeuralNetworksEx.h

index acbba23..d8c39e9 100644 (file)
@@ -227,6 +227,25 @@ typedef enum {
      *      A non-zero byte represents True, a hit. A zero indicates otherwise.
      */
     ANEURALNETWORKS_EQUAL_EX = 50011,
+
+    /**
+     * Computes element-wise absolute value of the input tensor.
+     *
+     * The output is calculated using this formula:
+     *
+     *      output = fabs(input)
+     *
+     * Supported tensor {@link OperandCode}:
+     * * {@link ANEURALNETWORKS_TENSOR_FLOAT32}
+     *
+     * Supported tensor rank: up to 4
+     *
+     * Inputs:
+     * * 0: An n-D tensor, specifying the first input.
+     *
+     * Outputs:
+     * * 0: The output tensor, of the same {@link OperandCode} and shape as input0.
+     */
     ANEURALNETWORKS_ABS_EX = 50012,
     /**
      * Packs a list of rank-R tensors into one rank- (R+1) tensor along the axis dimension.