ScaleLayer functions

Functions to acceleratе layer scale in Synet Framework. More...

Functions
SIMD_API void	SimdSynetScaleLayerForward (const float *src, const float *scale, const float *bias, size_t channels, size_t height, size_t width, float *dst, SimdTensorFormatType format, SimdSynetCompatibilityType compatibility)
	This function is used for forward propagation of ScaleLayer. More...
SIMD_API void *	SimdSynetScale8iInit (size_t batch, size_t channels, size_t spatial, SimdTensorDataType srcType, SimdTensorDataType dstType, SimdTensorFormatType format, SimdSynetCompatibilityType compatibility)
	Initilizes INT8 scale algorithm. More...
SIMD_API size_t	SimdSynetScale8iInternalBufferSize (const void *context)
	Gets size of internal buffer used inside INT8 scale algorithm. More...
SIMD_API void	SimdSynetScale8iSetParams (void *context, const float *scale, const float *bias, const float *const *stats)
	Sets scale, bias, parameters of activation function, input/output tensor statistics required for INT8 scale algorithm. More...
SIMD_API void	SimdSynetScale8iForward (void *context, const uint8_t *src, uint8_t *dst)
	Performs forward propagation of INT8 scale algorithm. More...

Detailed Description

Functions to acceleratе layer scale in Synet Framework.

Function Documentation

SimdSynetScaleLayerForward()

void SimdSynetScaleLayerForward	(	const float *	src,
		const float *	scale,
		const float *	bias,
		size_t	channels,
		size_t	height,
		size_t	width,
		float *	dst,
		SimdTensorFormatType	format,
		SimdSynetCompatibilityType	compatibility
	)

This function is used for forward propagation of ScaleLayer.

Algorithm's details (example for NCHW tensor format):

for(c = 0; c < channels; ++c)
    for(h = 0; h < height; ++h)
        for(w = 0; w < width; ++w)
            dst[(c*height + h)*width + w] = src[(c*height + h)*width + w]*scale[c] + (bias ? bias[c] : 0);

Note

This function is used in Synet Framework.

Parameters

[in]	src	- a pointer to the 32-bit float array with input image tensor. The size of the array is equal to channels * spatial.
[in]	scale	- a pointer to the 32-bit float array with scale coefficients. The size of the array is equal to channels.
[in]	bias	- a pointer to the 32-bit float array with bias coefficients. The size of the array is equal to channels. Can be NULL.
[in]	channels	- a number of channels in the (input/output) image tensor.
[in]	height	- a height of (input/output) image tensor.
[in]	width	- a width of (input/output) image tensor.
[out]	dst	- a pointer to the 32-bit float array with output image tensor. The size of the array is equal to channels * spatial.
[in]	format	- a format of (input/output) image tensor.
[in]	compatibility	- a flags of calculation compatibility.

SimdSynetScale8iInit()

void * SimdSynetScale8iInit	(	size_t	batch,
		size_t	channels,
		size_t	spatial,
		SimdTensorDataType	srcType,
		SimdTensorDataType	dstType,
		SimdTensorFormatType	format,
		SimdSynetCompatibilityType	compatibility
	)

Initilizes INT8 scale algorithm.

Parameters

[in]	batch	- a batch size.
[in]	channels	- a numbeo of channels.
[in]	spatial	- a spatial image size.
[in]	srcType	- an input data type (SimdTensorData32f or SimdTensorData8u).
[in]	dstType	- an output data type (SimdTensorData32f or SimdTensorData8u).
[in]	format	- a format of (input/output) image tensor.
[in]	compatibility	- a flags of calculation compatibility.

Returns

a pointer to INT8 scale context. On error it returns NULL. It must be released with using of function SimdRelease. This pointer is used in functions SimdSynetScale8iInternalBufferSize, SimdSynetScale8iSetParams and SimdSynetScale8iForward.

SimdSynetScale8iInternalBufferSize()

size_t SimdSynetScale8iInternalBufferSize

(

const void *

context

)

Gets size of internal buffer used inside INT8 scale algorithm.

Parameters

[in]

context

- a pointer to INT8 scale context. It must be created by function SimdSynetScale8iInit and released by function SimdRelease.

Returns

size of internal buffer used inside INT8 scale algorithm.

SimdSynetScale8iSetParams()

void SimdSynetScale8iSetParams	(	void *	context,
		const float *	scale,
		const float *	bias,
		const float *const *	stats
	)

Sets scale, bias, parameters of activation function, input/output tensor statistics required for INT8 scale algorithm.

Parameters

[in,out]	context	- a pointer to INT8 convolution context. It must be created by function SimdSynetScale8iInit and released by function SimdRelease.
[in]	scale	- a pointer to original (32-bit float point) scale.
[in]	bias	- a pointer to original (32-bit float point) bias. Can be NULL.
[in]	stats	- a pointer to pointers with statistics of input(min - stats[0], max - stats[1]) and output(min - stats[2], max - stats[3]) tensors. Can be NULL for subsequent calls of this function.

SimdSynetScale8iForward()

void SimdSynetScale8iForward	(	void *	context,
		const uint8_t *	src,
		uint8_t *	dst
	)

Performs forward propagation of INT8 scale algorithm.

Parameters

[in]	context	- a pointer to INT8 scale context. It must be created by function SimdSynetScale8iInit and released by function SimdRelease.
[in]	src	- a pointer to input tensor.
[out]	dst	- a pointer to output tensor.