arm_compute::NEQuantizationSymmetricKernel Class Reference

#include <NEQuantizationSymmetricKernel.h>

Collaboration diagram for arm_compute::NEQuantizationSymmetricKernel:

Public Member Functions
const char *	name () const override
	NEQuantizationSymmetricKernel ()
	NEQuantizationSymmetricKernel (const NEQuantizationSymmetricKernel &)=delete
NEQuantizationSymmetricKernel &	operator= (const NEQuantizationSymmetricKernel &)=delete
	NEQuantizationSymmetricKernel (NEQuantizationSymmetricKernel &&)=default
NEQuantizationSymmetricKernel &	operator= (NEQuantizationSymmetricKernel &&)=default
	~NEQuantizationSymmetricKernel ()=default
void	configure (const ITensor *input, ITensor *output, ITensor *scale_factor)
void	run (const Window &window, const ThreadInfo &info) override

Static Public Member Functions
static Status	validate (const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *scale_factor)

Detailed Description

Interface for the dequantization layer kernel.

Definition at line 51 of file NEQuantizationSymmetricKernel.h.

Constructor & Destructor Documentation

NEQuantizationSymmetricKernel() [1/3]

NEQuantizationSymmetricKernel::NEQuantizationSymmetricKernel

(

)

Default constructor

Definition at line 130 of file NEQuantizationSymmetricKernel.cpp.

  : _input(nullptr), _output(nullptr), _scale_factor(nullptr)
{
}

NEQuantizationSymmetricKernel() [2/3]

arm_compute::NEQuantizationSymmetricKernel::NEQuantizationSymmetricKernel ( const NEQuantizationSymmetricKernel &  )

delete

Prevent instances of this class from being copied (As this class contains pointers)

NEQuantizationSymmetricKernel() [3/3]

arm_compute::NEQuantizationSymmetricKernel::NEQuantizationSymmetricKernel ( NEQuantizationSymmetricKernel &&  )

default

Default Move Constructor.

~NEQuantizationSymmetricKernel()

arm_compute::NEQuantizationSymmetricKernel::~NEQuantizationSymmetricKernel ( )

default

Default destructor

References validate().

Member Function Documentation

configure()

void NEQuantizationSymmetricKernel::configure	(	const ITensor *	input,
		ITensor *	output,
		ITensor *	scale_factor
	)

Set input, output tensors.

Parameters

[in]	input	Source tensor. Data type supported: F16/F32.
[out]	output	Destination tensor with the same dimensions of input. Data type supported: S8.
[out]	scale_factor	Scale tensor of output. Data type supported: Same as input.

Definition at line 135 of file NEQuantizationSymmetricKernel.cpp.

{
  ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
  ARM_COMPUTE_ERROR_THROW_ON(
    validate_arguments(input->info(), output->info(), scale_factor->info()));
 
  _input = input;
  _output = output;
  _scale_factor = scale_factor;
 
  // Configure kernel window
  Window win_config = calculate_max_window(*input->info(), Steps());
 
  Coordinates coord;
  coord.set_num_dimensions(output->info()->num_dimensions());
  output->info()->set_valid_region(ValidRegion(coord, output->info()->tensor_shape()));
 
  INEKernel::configure(win_config);
}

Referenced by arm_compute::NEFullyConnectedHybridLayer::configure().

name()

const char * arm_compute::NEQuantizationSymmetricKernel::name ( ) const

inlineoverride

Definition at line 54 of file NEQuantizationSymmetricKernel.h.

{ return "NEQuantizationSymmetricKernel"; }

operator=() [1/2]

NEQuantizationSymmetricKernel & arm_compute::NEQuantizationSymmetricKernel::operator= ( const NEQuantizationSymmetricKernel &  )

delete

Prevent instances of this class from being copied (As this class contains pointers)

operator=() [2/2]

NEQuantizationSymmetricKernel & arm_compute::NEQuantizationSymmetricKernel::operator= ( NEQuantizationSymmetricKernel &&  )

default

Default move assignment operator

run()

void NEQuantizationSymmetricKernel::run ( const Window &  window, const ThreadInfo &  info  )

override

Definition at line 220 of file NEQuantizationSymmetricKernel.cpp.

{
  ARM_COMPUTE_UNUSED(info);
  ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
  ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
 
  switch (_input->info()->data_type())
  {
    case DataType::F32:
      NEQuantizationSymmetricKernel::quantize<float>(window);
      break;
#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
    case DataType::F16:
      NEQuantizationSymmetricKernel::quantize<float16_t>(window);
      break;
#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
    default:
      ARM_COMPUTE_ERROR("Unsupported data type.");
  }
}

References info.

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

validate()

Status NEQuantizationSymmetricKernel::validate ( const ITensorInfo *  input, const ITensorInfo *  output, const ITensorInfo *  scale_factor  )

static

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

Parameters

[in]	input	Input tensor info. Data types supported: F16/F32.
[in]	output	Output tensor info. Data types supported: S8.
[out]	scale_factor	Scale tensor of output. Data type supported: Same as input.

Returns

a status

Definition at line 156 of file NEQuantizationSymmetricKernel.cpp.

{
  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, scale_factor));
 
  return Status{};
}

Referenced by arm_compute::NEFullyConnectedHybridLayer::validate(), and ~NEQuantizationSymmetricKernel().

---

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

• compute/ARMComputeEx/arm_compute/core/NEON/kernels/NEQuantizationSymmetricKernel.h
• compute/ARMComputeEx/src/core/NEON/kernels/NEQuantizationSymmetricKernel.cpp