diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2017-08-18 10:16:09 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:35:24 +0000 |
commit | 409ee0a69799364797263d13dd95936c851bfe80 (patch) | |
tree | 297e396b46df7f8079173ba4ccd6f7fb2aad560d /arm_compute/core/NEON/kernels | |
parent | d763cfbc972cded289a2402a6238416d371bdf33 (diff) | |
download | ComputeLibrary-409ee0a69799364797263d13dd95936c851bfe80.tar.gz |
COMPMID-417: Add in-place support for batch-normalization.
Change-Id: I4b0c9348f3bc2addc198a76fadd1b583abf42b60
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/84434
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Diffstat (limited to 'arm_compute/core/NEON/kernels')
-rw-r--r-- | arm_compute/core/NEON/kernels/NEBatchNormalizationLayerKernel.h | 25 |
1 files changed, 14 insertions, 11 deletions
diff --git a/arm_compute/core/NEON/kernels/NEBatchNormalizationLayerKernel.h b/arm_compute/core/NEON/kernels/NEBatchNormalizationLayerKernel.h index 29fcbd26a0..8ac70be727 100644 --- a/arm_compute/core/NEON/kernels/NEBatchNormalizationLayerKernel.h +++ b/arm_compute/core/NEON/kernels/NEBatchNormalizationLayerKernel.h @@ -49,24 +49,27 @@ public: ~NEBatchNormalizationLayerKernel() = default; /** Set the input and output tensors. * - * @param[in] input Source tensor. 3 lower dimensions represent a single input with dimensions [width, height, FM]. - * The rest are optional and used for representing batches. Data types supported: QS8/F32. - * @param[in] mean Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input - * @param[in] var Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input - * @param[in] gamma Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input - * @param[in] beta Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input - * @param[in] epsilon Small value to avoid division with zero. - * @param[out] output Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input + * @note If the output tensor is a nullptr, the batch normalization function will be performed in-place + * + * @param[in, out] input Source tensor. In case of @p output tensor = nullptr, this tensor will store the result. + * 3 lower dimensions represent a single input with dimensions [width, height, FM]. + * The rest are optional and used for representing batches. Data types supported: QS8/QS16/F16/F32. + * @param[in] mean Mean values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input + * @param[in] var Variance values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input + * @param[in] gamma Gamma values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input + * @param[in] beta Beta values tensor. 1 dimension with size equal to the feature maps [FM]. Data types supported: Same as @p input + * @param[in] epsilon Small value to avoid division with zero. + * @param[out] output Destination tensor. Output will have the same number of dimensions as input. Data type supported: same as @p input */ - void configure(const ITensor *input, ITensor *output, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon); + void configure(ITensor *input, ITensor *output, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon); // Inherited methods overridden: void run(const Window &window) override; private: - using BatchNormFunction = void(const ITensor *input, ITensor *output, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon, const Window &window); + using BatchNormFunction = void(ITensor *input, ITensor *output, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon, const Window &window); BatchNormFunction *_func; - const ITensor *_input; + ITensor *_input; ITensor *_output; const ITensor *_mean; const ITensor *_var; |