arm_compute::CLTransposeConvLayer Class Reference

#include <CLTransposeConvLayer.h>

Collaboration diagram for arm_compute::CLTransposeConvLayer:

Public Member Functions
	CLTransposeConvLayer (std::shared_ptr< IMemoryManager > memory_manager=nullptr)
void	configure (ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &deconv_info, unsigned int invalid_right, unsigned int invalid_bottom, const WeightsInfo &weights_info=WeightsInfo())
void	configure (const CLCompileContext &compile_context, ICLTensor *input, ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &deconv_info, unsigned int invalid_right, unsigned int invalid_bottom, const WeightsInfo &weights_info=WeightsInfo())
void	run () override
void	prepare () override

Static Public Member Functions
static Status	validate (const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &deconv_info, unsigned int invalid_right, unsigned int invalid_bottom, const WeightsInfo &weights_info=WeightsInfo())
static DeconvolutionMethod	get_deconvolution_method (const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &deconv_info, unsigned int invalid_right, unsigned int invalid_bottom, const WeightsInfo &weights_info)

Detailed Description

Basic function to compute the deconvolution layer. This function calls the following OpenCL kernels/functions:

1. CLGEMMDeconvolutionLayer
2. CLDirectTransposeConvLayer

Definition at line 58 of file CLTransposeConvLayer.h.

Constructor & Destructor Documentation

CLTransposeConvLayer()

CLTransposeConvLayer::CLTransposeConvLayer

(

std::shared_ptr< IMemoryManager >

memory_manager = nullptr

)

Default constructor

Definition at line 56 of file CLTransposeConvLayer.cpp.

  : _memory_manager(std::move(memory_manager)), _function()
{
}

Member Function Documentation

configure() [1/2]

void CLTransposeConvLayer::configure	(	const CLCompileContext &	compile_context,
		ICLTensor *	input,
		ICLTensor *	weights,
		const ICLTensor *	bias,
		ICLTensor *	output,
		const PadStrideInfo &	deconv_info,
		unsigned int	invalid_right,
		unsigned int	invalid_bottom,
		const WeightsInfo &	weights_info = WeightsInfo()
	)

Set the input, weights, biases and output tensors.

Parameters

[in]	compile_context	The compile context to be used.
[in,out]	input	Input tensor. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: QASYMM8_SIGNED/QASYMM8/F16/F32.
[in]	weights	The 4d weights with dimensions [width, height, IFM, OFM]. Data type supported: Same as input.
[in]	bias	(Optional) The biases have one dimension. Data type supported: Same as input.
[out]	output	Output tensor. The output has the same number of dimensions as the input.
[in]	deconv_info	Contains padding and policies to be used in the deconvolution, this is described in PadStrideInfo.
[in]	invalid_right	The number of zeros added to right edge of the output.
[in]	invalid_bottom	The number of zeros added to bottom edge of the output.
[in]	weights_info	(Optional) Weights information needed for CLConvolutionLayer, specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel.

Definition at line 70 of file CLTransposeConvLayer.cpp.

{
  ARM_COMPUTE_ERROR_ON_NULLPTR(input, weights, output);
 
  switch (CLTransposeConvLayer::get_deconvolution_method(input->info(), weights->info(), nullptr,
                                                         output->info(), deconv_info, invalid_right,
                                                         invalid_bottom, weights_info))
  {
    case DeconvolutionMethod::DIRECT:
    {
      auto f = std::make_unique<CLDirectTransposeConvLayer>();
      f->configure(compile_context, input, weights, bias, output, deconv_info, invalid_right,
                   invalid_bottom, weights_info);
      _function = std::move(f);
      break;
    }
    case DeconvolutionMethod::GEMM:
    {
      auto f = std::make_unique<CLGEMMDeconvolutionLayer>(_memory_manager);
      f->configure(compile_context, input, weights, bias, output, deconv_info);
      _function = std::move(f);
      break;
    }
    default:
      ARM_COMPUTE_ERROR("Not supported.");
      break;
  }
}

References get_deconvolution_method().

configure() [2/2]

void CLTransposeConvLayer::configure	(	ICLTensor *	input,
		ICLTensor *	weights,
		const ICLTensor *	bias,
		ICLTensor *	output,
		const PadStrideInfo &	deconv_info,
		unsigned int	invalid_right,
		unsigned int	invalid_bottom,
		const WeightsInfo &	weights_info = WeightsInfo()
	)

Set the input, weights, biases and output tensors.

Parameters

[in,out]	input	Input tensor. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: QASYMM8_SIGNED/QASYMM8/F16/F32.
[in]	weights	The 4d weights with dimensions [width, height, IFM, OFM]. Data type supported: Same as input.
[in]	bias	(Optional) The biases have one dimension. Data type supported: Same as input.
[out]	output	Output tensor. The output has the same number of dimensions as the input.
[in]	deconv_info	Contains padding and policies to be used in the deconvolution, this is described in PadStrideInfo.
[in]	invalid_right	The number of zeros added to right edge of the output.
[in]	invalid_bottom	The number of zeros added to bottom edge of the output.
[in]	weights_info	(Optional) Weights information needed for CLConvolutionLayer, specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel.

Definition at line 61 of file CLTransposeConvLayer.cpp.

{
  configure(CLKernelLibrary::get().get_compile_context(), input, weights, bias, output, deconv_info,
            invalid_right, invalid_bottom, weights_info);
}

References configure().

Referenced by configure().

get_deconvolution_method()

DeconvolutionMethod CLTransposeConvLayer::get_deconvolution_method ( const ITensorInfo *  input, const ITensorInfo *  weights, const ITensorInfo *  bias, ITensorInfo *  output, const PadStrideInfo &  deconv_info, unsigned int  invalid_right, unsigned int  invalid_bottom, const WeightsInfo &  weights_info  )

static

Definition at line 133 of file CLTransposeConvLayer.cpp.

{
  ARM_COMPUTE_UNUSED(output, bias, weights_info);
 
  const DataLayout data_layout = input->data_layout();
 
  const size_t idx_w = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
  const size_t idx_h = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
 
  if (weights->dimension(idx_w) != deconv_info.stride().first ||
      weights->dimension(idx_h) != deconv_info.stride().second || invalid_right != 0 ||
      invalid_bottom != 0)
  {
    return DeconvolutionMethod::DIRECT;
  }
 
  return DeconvolutionMethod::GEMM;
}

Referenced by configure(), and validate().

prepare()

void CLTransposeConvLayer::prepare ( )

override

Definition at line 161 of file CLTransposeConvLayer.cpp.

{ _function->prepare(); }

Referenced by run().

run()

void CLTransposeConvLayer::run ( )

override

Definition at line 155 of file CLTransposeConvLayer.cpp.

{
  prepare();
  _function->run();
}

References prepare().

Referenced by package.infer.session::inference().

validate()

Status CLTransposeConvLayer::validate ( const ITensorInfo *  input, const ITensorInfo *  weights, const ITensorInfo *  bias, ITensorInfo *  output, const PadStrideInfo &  deconv_info, unsigned int  invalid_right, unsigned int  invalid_bottom, const WeightsInfo &  weights_info = WeightsInfo()  )

static

Static function to check if given info will lead to a valid configuration of CLTransposeConvLayer

Parameters

[in]	input	Input tensor info. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: QASYMM8_SIGNED/QASYMM8/F16/F32.
[in]	weights	The 4d weights info with dimensions [width, height, IFM, OFM]. Data type supported: Same as input.
[in]	bias	(Optional) The biases have one dimension. Data type supported: Same as input.
[in]	output	Output tensor info. The output has the same number of dimensions as the input.
[in]	deconv_info	Contains padding and policies to be used in the deconvolution, this is described in PadStrideInfo.
[in]	invalid_right	The number of zeros added to right edge of the output.
[in]	invalid_bottom	The number of zeros added to bottom edge of the output.
[in]	weights_info	(Optional) Weights information needed for CLConvolutionLayer, specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel.

Returns

a status

Definition at line 102 of file CLTransposeConvLayer.cpp.

{
  ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, weights, output);
  switch (CLTransposeConvLayer::get_deconvolution_method(
    input, weights, bias, output, deconv_info, invalid_right, invalid_bottom, weights_info))
  {
    case DeconvolutionMethod::DIRECT:
    {
      // Validate direct convolution layer
      ARM_COMPUTE_RETURN_ON_ERROR(CLDirectTransposeConvLayer::validate(
        input, weights, bias, output, deconv_info, invalid_right, invalid_bottom, weights_info));
      break;
    }
    case DeconvolutionMethod::GEMM:
    {
      // Validate gemm-based convolution layer
      ARM_COMPUTE_RETURN_ON_ERROR(
        CLGEMMDeconvolutionLayer::validate(input, weights, bias, output, deconv_info));
      break;
    }
    default:
      ARM_COMPUTE_ERROR("Not supported.");
      break;
  }
 
  return Status{};
}

References get_deconvolution_method(), and arm_compute::CLDirectTransposeConvLayer::validate().

---

The documentation for this class was generated from the following files:

• compute/ARMComputeEx/arm_compute/runtime/CL/functions/CLTransposeConvLayer.h
• compute/ARMComputeEx/src/runtime/CL/functions/CLTransposeConvLayer.cpp