diff options
Diffstat (limited to 'src/core')
-rw-r--r-- | src/core/NEON/NEKernels.h | 1 | ||||
-rw-r--r-- | src/core/NEON/kernels/NEWeightsReshapeKernel.cpp | 177 | ||||
-rw-r--r-- | src/core/cpu/kernels/CpuIm2ColKernel.cpp | 2 | ||||
-rw-r--r-- | src/core/cpu/kernels/CpuIm2ColKernel.h | 2 | ||||
-rw-r--r-- | src/core/cpu/kernels/CpuWeightsReshapeKernel.cpp | 170 | ||||
-rw-r--r-- | src/core/cpu/kernels/CpuWeightsReshapeKernel.h (renamed from src/core/NEON/kernels/NEWeightsReshapeKernel.h) | 66 |
6 files changed, 196 insertions, 222 deletions
diff --git a/src/core/NEON/NEKernels.h b/src/core/NEON/NEKernels.h index 6c6c51dd87..6d45a9d80c 100644 --- a/src/core/NEON/NEKernels.h +++ b/src/core/NEON/NEKernels.h @@ -65,6 +65,5 @@ #include "src/core/NEON/kernels/NEStackLayerKernel.h" #include "src/core/NEON/kernels/NEStridedSliceKernel.h" #include "src/core/NEON/kernels/NETileKernel.h" -#include "src/core/NEON/kernels/NEWeightsReshapeKernel.h" #endif /* ARM_COMPUTE_NEKERNELS_H */ diff --git a/src/core/NEON/kernels/NEWeightsReshapeKernel.cpp b/src/core/NEON/kernels/NEWeightsReshapeKernel.cpp deleted file mode 100644 index 9bef9c30d9..0000000000 --- a/src/core/NEON/kernels/NEWeightsReshapeKernel.cpp +++ /dev/null @@ -1,177 +0,0 @@ -/* - * Copyright (c) 2017-2021 Arm Limited. - * - * SPDX-License-Identifier: MIT - * - * Permission is hereby granted, free of charge, to any person obtaining a copy - * of this software and associated documentation files (the "Software"), to - * deal in the Software without restriction, including without limitation the - * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or - * sell copies of the Software, and to permit persons to whom the Software is - * furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in all - * copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE - * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, - * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE - * SOFTWARE. - */ -#include "src/core/NEON/kernels/NEWeightsReshapeKernel.h" - -#include "arm_compute/core/Helpers.h" -#include "arm_compute/core/Validate.h" -#include "src/core/helpers/AutoConfiguration.h" -#include "src/core/helpers/WindowHelpers.h" - -namespace arm_compute -{ -namespace -{ -TensorShape get_output_shape(const ITensorInfo *input, bool has_bias) -{ - TensorShape output_shape{ input->tensor_shape() }; - - output_shape.collapse(3); - const size_t tmp_dim = output_shape[0]; - output_shape.set(0, output_shape[1]); - output_shape.set(1, tmp_dim + (has_bias ? 1 : 0)); - - return output_shape; -} - -Status validate_arguments(const ITensorInfo *input, const ITensorInfo *biases, const ITensorInfo *output) -{ - ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output); - //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use CPU FP16 instructions. - ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN); - - if(biases != nullptr) - { - ARM_COMPUTE_RETURN_ERROR_ON(is_data_type_quantized_asymmetric(input->data_type())); - ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, biases); - ARM_COMPUTE_RETURN_ERROR_ON((input->num_dimensions() == 4) && (biases->num_dimensions() != 1)); - ARM_COMPUTE_RETURN_ERROR_ON((input->num_dimensions() == 5) && (biases->num_dimensions() != 2)); - ARM_COMPUTE_RETURN_ERROR_ON((input->num_dimensions() == 4) && (biases->dimension(0) != input->tensor_shape()[3])); - ARM_COMPUTE_RETURN_ERROR_ON((input->num_dimensions() == 5) && (biases->dimension(0) != input->tensor_shape()[3] || biases->dimension(1) != input->tensor_shape()[4])); - } - - // Checks performed when output is configured - if(output->total_size() != 0) - { - ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), get_output_shape(input, biases != nullptr)); - ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); - ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output); - } - - return Status{}; -} - -std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input) -{ - Window window = calculate_max_window(*input, Steps()); - window.set(Window::DimX, Window::Dimension(0, input->dimension(0), input->dimension(0))); - window.set(Window::DimY, Window::Dimension(0, input->dimension(1), input->dimension(1))); - window.set(Window::DimZ, Window::Dimension(0, input->dimension(2), input->dimension(2))); - - // The NEConvolutionLayerWeightsReshapeKernel doesn't need padding so update_window_and_padding() can be skipped - - return std::make_pair(Status{}, window); -} -} // namespace - -NEWeightsReshapeKernel::NEWeightsReshapeKernel() - : _input(nullptr), _bias(nullptr), _output(nullptr) -{ -} - -void NEWeightsReshapeKernel::configure(const ITensor *input, const ITensor *bias, ITensor *output) -{ - ARM_COMPUTE_ERROR_ON_NULLPTR(input, output); - - // Output tensor auto inizialitation if not yet initialized - auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(get_output_shape(input->info(), (bias != nullptr)))); - - // Perform validation step - ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), - (bias != nullptr) ? bias->info() : nullptr, - output->info())); - - _input = input; - _bias = bias; - _output = output; - - // Configure kernel - auto win_config = validate_and_configure_window(input->info()); - ARM_COMPUTE_ERROR_THROW_ON(win_config.first); - INEKernel::configure(win_config.second); -} - -Status NEWeightsReshapeKernel::validate(const ITensorInfo *input, const ITensorInfo *biases, const ITensorInfo *output) -{ - ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, biases, output)); - ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get()).first); - - return Status{}; -} - -void NEWeightsReshapeKernel::run(const Window &window, const ThreadInfo &info) -{ - ARM_COMPUTE_UNUSED(info); - ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); - ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window); - - const unsigned int kernel_size_x = _input->info()->dimension(0); - const unsigned int kernel_size_y = _input->info()->dimension(1); - const unsigned int kernel_depth = _input->info()->dimension(2); - const unsigned int input_stride_x = _input->info()->strides_in_bytes().x(); - const unsigned int input_stride_y = _input->info()->strides_in_bytes().y(); - const unsigned int input_stride_z = _input->info()->strides_in_bytes().z(); - const unsigned int output_stride_y = _output->info()->strides_in_bytes().y(); - - // Create iterators - Iterator in(_input, window); - execute_window_loop(window, [&](const Coordinates & id) - { - // Get column index - const int kernel_idx = id[3]; - const int kernel_idz = id[4]; - - // Setup pointers - const uint8_t *tmp_input_ptr = in.ptr(); - uint8_t *tmp_output_ptr = _output->ptr_to_element(Coordinates(kernel_idx, 0, kernel_idz)); - const uint8_t *curr_input_row_ptr = tmp_input_ptr; - const uint8_t *curr_input_depth_ptr = tmp_input_ptr; - - // Linearize volume - for(unsigned int d = 0; d < kernel_depth; ++d) - { - for(unsigned int j = 0; j < kernel_size_y; ++j) - { - for(unsigned int i = 0; i < kernel_size_x; ++i) - { - std::memcpy(tmp_output_ptr, tmp_input_ptr, _input->info()->element_size()); - tmp_input_ptr += input_stride_x; - tmp_output_ptr += output_stride_y; - } - curr_input_row_ptr += input_stride_y; - tmp_input_ptr = curr_input_row_ptr; - } - curr_input_depth_ptr += input_stride_z; - curr_input_row_ptr = curr_input_depth_ptr; - tmp_input_ptr = curr_input_depth_ptr; - } - - // Add bias - if(_bias != nullptr) - { - std::memcpy(tmp_output_ptr, _bias->ptr_to_element(Coordinates(kernel_idx, kernel_idz)), _input->info()->element_size()); - } - }, - in); -} -} // namespace arm_compute diff --git a/src/core/cpu/kernels/CpuIm2ColKernel.cpp b/src/core/cpu/kernels/CpuIm2ColKernel.cpp index a5dbcc29c8..ca6c9bfab4 100644 --- a/src/core/cpu/kernels/CpuIm2ColKernel.cpp +++ b/src/core/cpu/kernels/CpuIm2ColKernel.cpp @@ -331,7 +331,7 @@ void CpuIm2ColKernel::run_im2col(const ITensor *src, ITensor *dst, const Window in, out); } -void CpuIm2ColKernel::configure(ITensorInfo *src, ITensorInfo *dst, const Size2D &kernel_dims, const PadStrideInfo &conv_info, +void CpuIm2ColKernel::configure(const ITensorInfo *src, ITensorInfo *dst, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias, const Size2D &dilation, unsigned int num_groups) { ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst); diff --git a/src/core/cpu/kernels/CpuIm2ColKernel.h b/src/core/cpu/kernels/CpuIm2ColKernel.h index 4301a237fe..ffac5077b2 100644 --- a/src/core/cpu/kernels/CpuIm2ColKernel.h +++ b/src/core/cpu/kernels/CpuIm2ColKernel.h @@ -77,7 +77,7 @@ public: * @param[in] dilation (Optional) Dilation, in elements, across x and y. Defaults to (1, 1). * @param[in] num_groups (Optional) Number of groups when performing a grouped convolution. num_groups != 1 is not supported */ - void configure(ITensorInfo *src, ITensorInfo *dst, const Size2D &kernel_dims, const PadStrideInfo &conv_info, + void configure(const ITensorInfo *src, ITensorInfo *dst, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias, const Size2D &dilation = Size2D(1U, 1U), unsigned int num_groups = 1); /** Static function to check if given info will lead to a valid configuration * diff --git a/src/core/cpu/kernels/CpuWeightsReshapeKernel.cpp b/src/core/cpu/kernels/CpuWeightsReshapeKernel.cpp new file mode 100644 index 0000000000..79f058944d --- /dev/null +++ b/src/core/cpu/kernels/CpuWeightsReshapeKernel.cpp @@ -0,0 +1,170 @@ +/* + * Copyright (c) 2017-2021 Arm Limited. + * + * SPDX-License-Identifier: MIT + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#include "src/core/cpu/kernels/CpuWeightsReshapeKernel.h" + +#include "arm_compute/core/Helpers.h" +#include "arm_compute/core/Validate.h" +#include "src/core/helpers/AutoConfiguration.h" +#include "src/core/helpers/WindowHelpers.h" + +namespace arm_compute +{ +namespace cpu +{ +namespace kernels +{ +namespace +{ +TensorShape get_output_shape(const ITensorInfo *src, bool has_bias) +{ + TensorShape output_shape{ src->tensor_shape() }; + + output_shape.collapse(3); + const size_t tmp_dim = output_shape[0]; + output_shape.set(0, output_shape[1]); + output_shape.set(1, tmp_dim + (has_bias ? 1 : 0)); + + return output_shape; +} + +Status validate_arguments(const ITensorInfo *src, const ITensorInfo *biases, const ITensorInfo *dst) +{ + ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, dst); + //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(src) is not needed here as this kernel doesn't use CPU FP16 instructions. + ARM_COMPUTE_RETURN_ERROR_ON(src->data_type() == DataType::UNKNOWN); + + if(biases != nullptr) + { + ARM_COMPUTE_RETURN_ERROR_ON(is_data_type_quantized_asymmetric(src->data_type())); + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, biases); + ARM_COMPUTE_RETURN_ERROR_ON((src->num_dimensions() == 4) && (biases->num_dimensions() != 1)); + ARM_COMPUTE_RETURN_ERROR_ON((src->num_dimensions() == 5) && (biases->num_dimensions() != 2)); + ARM_COMPUTE_RETURN_ERROR_ON((src->num_dimensions() == 4) && (biases->dimension(0) != src->tensor_shape()[3])); + ARM_COMPUTE_RETURN_ERROR_ON((src->num_dimensions() == 5) && (biases->dimension(0) != src->tensor_shape()[3] || biases->dimension(1) != src->tensor_shape()[4])); + } + + // Checks performed when output is configured + if(dst->total_size() != 0) + { + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(dst->tensor_shape(), get_output_shape(src, biases != nullptr)); + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst); + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(src, dst); + } + + return Status{}; +} +} // namespace + +void CpuWeightsReshapeKernel::configure(const ITensorInfo *src, const ITensorInfo *biases, ITensorInfo *dst) +{ + ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst); + + // Output tensor auto inizialitation if not yet initialized + auto_init_if_empty(*dst, src->clone()->set_tensor_shape(get_output_shape(src, (biases != nullptr)))); + + // Perform validation step + ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, + biases, + dst)); + + // Configure kernel + Window window = calculate_max_window(*src, Steps()); + window.set(Window::DimX, Window::Dimension(0, src->dimension(0), src->dimension(0))); + window.set(Window::DimY, Window::Dimension(0, src->dimension(1), src->dimension(1))); + window.set(Window::DimZ, Window::Dimension(0, src->dimension(2), src->dimension(2))); + ICpuKernel::configure(window); +} + +Status CpuWeightsReshapeKernel::validate(const ITensorInfo *src, const ITensorInfo *biases, const ITensorInfo *dst) +{ + ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, biases, dst)); + return Status{}; +} + +void CpuWeightsReshapeKernel::run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) +{ + ARM_COMPUTE_UNUSED(info); + ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); + ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICpuKernel::window(), window); + + auto src = tensors.get_const_tensor(TensorType::ACL_SRC); + auto biases = tensors.get_const_tensor(TensorType::ACL_BIAS); + auto dst = tensors.get_tensor(TensorType::ACL_DST); + + const unsigned int kernel_size_x = src->info()->dimension(0); + const unsigned int kernel_size_y = src->info()->dimension(1); + const unsigned int kernel_depth = src->info()->dimension(2); + const unsigned int input_stride_x = src->info()->strides_in_bytes().x(); + const unsigned int input_stride_y = src->info()->strides_in_bytes().y(); + const unsigned int input_stride_z = src->info()->strides_in_bytes().z(); + const unsigned int output_stride_y = dst->info()->strides_in_bytes().y(); + + // Create iterators + Iterator in(src, window); + execute_window_loop(window, [&](const Coordinates & id) + { + // Get column index + const int kernel_idx = id[3]; + const int kernel_idz = id[4]; + + // Setup pointers + const uint8_t *tmp_input_ptr = in.ptr(); + uint8_t *tmp_output_ptr = dst->ptr_to_element(Coordinates(kernel_idx, 0, kernel_idz)); + const uint8_t *curr_input_row_ptr = tmp_input_ptr; + const uint8_t *curr_input_depth_ptr = tmp_input_ptr; + + // Linearize volume + for(unsigned int d = 0; d < kernel_depth; ++d) + { + for(unsigned int j = 0; j < kernel_size_y; ++j) + { + for(unsigned int i = 0; i < kernel_size_x; ++i) + { + std::memcpy(tmp_output_ptr, tmp_input_ptr, src->info()->element_size()); + tmp_input_ptr += input_stride_x; + tmp_output_ptr += output_stride_y; + } + curr_input_row_ptr += input_stride_y; + tmp_input_ptr = curr_input_row_ptr; + } + curr_input_depth_ptr += input_stride_z; + curr_input_row_ptr = curr_input_depth_ptr; + tmp_input_ptr = curr_input_depth_ptr; + } + + // Add bias + if(biases != nullptr) + { + std::memcpy(tmp_output_ptr, biases->ptr_to_element(Coordinates(kernel_idx, kernel_idz)), src->info()->element_size()); + } + }, + in); +} +const char *CpuWeightsReshapeKernel::name() const +{ + return "CpuWeightsReshapeKernel"; +} +} // namespace kernels +} // namespace cpu +} // namespace arm_compute
\ No newline at end of file diff --git a/src/core/NEON/kernels/NEWeightsReshapeKernel.h b/src/core/cpu/kernels/CpuWeightsReshapeKernel.h index 5701c84cac..eea150a96e 100644 --- a/src/core/NEON/kernels/NEWeightsReshapeKernel.h +++ b/src/core/cpu/kernels/CpuWeightsReshapeKernel.h @@ -21,15 +21,18 @@ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -#ifndef ARM_COMPUTE_NEWEIGHTSRESHAPEKERNEL_H -#define ARM_COMPUTE_NEWEIGHTSRESHAPEKERNEL_H +#ifndef ARM_COMPUTE_CPU_WEIGHTSRESHAPE_KERNEL_H +#define ARM_COMPUTE_CPU_WEIGHTSRESHAPE_KERNEL_H -#include "src/core/NEON/INEKernel.h" +#include "src/core/common/Macros.h" +#include "src/core/cpu/ICpuKernel.h" namespace arm_compute { -class ITensor; - +namespace cpu +{ +namespace kernels +{ /** Kernel to perform reshaping on the weights used by convolution and locally connected layer * * Rearranges each 3-dimensional kernel to a single row leading to a matrix with linearized kernels. @@ -53,57 +56,36 @@ class ITensor; * \end{array} \right) * @f] */ -class NEWeightsReshapeKernel : public INEKernel +class CpuWeightsReshapeKernel : public ICpuKernel { public: - const char *name() const override - { - return "NEWeightsReshapeKernel"; - } - /** Constructor.*/ - NEWeightsReshapeKernel(); - /** Prevent instances of this class from being copied (As this class contains pointers) */ - NEWeightsReshapeKernel(const NEWeightsReshapeKernel &) = delete; - /** Prevent instances of this class from being copied (As this class contains pointers) */ - NEWeightsReshapeKernel &operator=(const NEWeightsReshapeKernel &) = delete; - /** Allow instances of this class to be moved */ - NEWeightsReshapeKernel(NEWeightsReshapeKernel &&) = default; - /** Allow instances of this class to be moved */ - NEWeightsReshapeKernel &operator=(NEWeightsReshapeKernel &&) = default; - /** Default destructor */ - ~NEWeightsReshapeKernel() = default; + /** Default constructor */ + CpuWeightsReshapeKernel() = default; + ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuWeightsReshapeKernel); /** Set the input and output of the kernel. * - * @param[in] input The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared, + * @param[in] src The input tensor info to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared, * and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM, num_patches] if unshared. * Data types supported: All - * @param[in] bias The shared biases tensor to append. Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with + * @param[in] biases The shared biases tensor info to append. Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with * dimensions [OFM, num_patches] if unshared. Data types supported: Same as @p input * @warning Appending biases to weights reshaped matrix is not supported for quantized asymmetric types. - * @param[out] output The output tensor. Data types supported: Same as @p input + * @param[out] dst The output tensor info. Data types supported: Same as @p src */ - void configure(const ITensor *input, const ITensor *bias, ITensor *output); - /** Static function to check if given info will lead to a valid configuration of @ref NEWeightsReshapeKernel + void configure(const ITensorInfo *src, const ITensorInfo *biases, ITensorInfo *dst); + /** Static function to check if given info will lead to a valid configuration * - * @param[in] input The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared, - * and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM, num_patches] if unshared. - * Data types supported: All - * @param[in] biases The shared biases tensor to append. Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with - * dimensions [OFM, num_patches] if unshared. Data types supported: Same as @p input - * @warning Appending biases to weights reshaped matrix is not supported for quantized asymmetric types. - * @param[in] output The output tensor. Should be a 2D Tensor. Data types supported: Same as @p input + * Similar to CpuWeightsReshapeKernel::configure() * * @return a status */ - static Status validate(const ITensorInfo *input, const ITensorInfo *biases, const ITensorInfo *output); + static Status validate(const ITensorInfo *src, const ITensorInfo *biases, const ITensorInfo *dst); // Inherited methods overridden: - void run(const Window &window, const ThreadInfo &info) override; - -private: - const ITensor *_input; - const ITensor *_bias; - ITensor *_output; + void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override; + const char *name() const override; }; +} // namespace kernels +} // namespace cpu } // namespace arm_compute -#endif /*ARM_COMPUTE_NEWEIGHTSRESHAPEKERNEL_H */ +#endif /* ARM_COMPUTE_CPU_WEIGHTSRESHAPE_KERNEL_H */ |