diff options
| author | Georgios Pinitas <georgios.pinitas@arm.com> | 2017-10-23 20:29:30 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:35:24 +0000 |
| commit | 81a26ad6b626ce2da83659d7c6c17b6104d1f203 (patch) | |
| tree | 536807115771f1a4b06048892d1d4e17c98779de /arm_compute/core | |
| parent | 511347a7282b948bddfc071e9a8fa08e79da25b4 (diff) | |
| download | ComputeLibrary-81a26ad6b626ce2da83659d7c6c17b6104d1f203.tar.gz | |
COMPMID-643: Add bias to CLDepthwiseConvolution.
Change-Id: Ibfe7b8c1172d10cbcae7971fe86b82090519d31d
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/92798
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Jaroslaw Rzepecki <jaroslaw.rzepecki@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Diffstat (limited to 'arm_compute/core')
3 files changed, 10 insertions, 4 deletions
diff --git a/arm_compute/core/CL/kernels/CLDepthwiseConvolution3x3Kernel.h b/arm_compute/core/CL/kernels/CLDepthwiseConvolution3x3Kernel.h index 4e69f551b8..b3d1eaf3af 100644 --- a/arm_compute/core/CL/kernels/CLDepthwiseConvolution3x3Kernel.h +++ b/arm_compute/core/CL/kernels/CLDepthwiseConvolution3x3Kernel.h @@ -49,10 +49,12 @@ public: * * @param[in] input Source tensor. DataType supported: F32. * @param[out] output Destination tensor. Data type supported: Same as @p input. - * @param[in] weights Weights tensor. These are 3D tensors with dimensions [3, 3, IFM]. Data type supported: Same as @p input. + * @param[in] weights Weights tensor. A 3D tensor with dimensions [3, 3, IFM]. Data type supported: Same as @p input. + * @param[in] biases (Optional) Biases tensor. A 1D tensor with dimensions [IFM]. Must be nullptr if not needed. + * Data type supported: Same as @p input. * @param[in] conv_info Padding and stride information to use for the convolution. */ - void configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *weights, const PadStrideInfo &conv_info); + void configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *weights, const ICLTensor *biases, const PadStrideInfo &conv_info); // Inherited methods overridden: void run(const Window &window, cl::CommandQueue &queue) override; @@ -63,6 +65,7 @@ private: const ICLTensor *_input; ICLTensor *_output; const ICLTensor *_weights; + const ICLTensor *_biases; unsigned int _conv_stride_x; unsigned int _conv_stride_y; unsigned int _conv_pad_x; diff --git a/arm_compute/core/CL/kernels/CLDepthwiseIm2ColKernel.h b/arm_compute/core/CL/kernels/CLDepthwiseIm2ColKernel.h index ae56adfa30..7e786e8df8 100644 --- a/arm_compute/core/CL/kernels/CLDepthwiseIm2ColKernel.h +++ b/arm_compute/core/CL/kernels/CLDepthwiseIm2ColKernel.h @@ -56,8 +56,9 @@ public: * while every dimension above 3 represents a batch. Data types supported: Same as @p input * @param[in] kernel_dims The kernel dimensions (width and height). * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo. + * @param[in] has_bias Boolean that specifies if the depthwise convolution has bias. */ - void configure(const ICLTensor *input, ICLTensor *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info); + void configure(const ICLTensor *input, ICLTensor *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias = false); // Inherited methods overridden: void run(const Window &window, cl::CommandQueue &queue) override; diff --git a/arm_compute/core/CL/kernels/CLDepthwiseWeightsReshapeKernel.h b/arm_compute/core/CL/kernels/CLDepthwiseWeightsReshapeKernel.h index d493d9f052..7989257d34 100644 --- a/arm_compute/core/CL/kernels/CLDepthwiseWeightsReshapeKernel.h +++ b/arm_compute/core/CL/kernels/CLDepthwiseWeightsReshapeKernel.h @@ -52,14 +52,16 @@ public: * * @param[in] input The input tensor to convert. 3 lower dimensions represent a single input [width, height, IFM]. Data type supported: F32. * @param[out] output The output tensor. Data type supported: same as @p input. + * @param[in] biases (Optional) The input biases to add. Shape [IFM]. Data type supported: same as @p input. */ - void configure(const ICLTensor *input, ICLTensor *output); + void configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *biases = nullptr); // Inherited methods overridden: void run(const Window &window, cl::CommandQueue &queue) override; private: const ICLTensor *_input; + const ICLTensor *_biases; ICLTensor *_output; }; } // arm_compute |