odla_ops_quantization.h File Reference

Go to the source code of this file.

Data Structures
struct	odla_value_quant_info
	Quantization info for each odla value. More...

Functions
ODLA_API_EXPORT odla_value ODLA_API_CALL	odla_Dequantize (odla_value input, odla_value scale, odla_value zero_point, odla_int32 axis, odla_element_type target_data_type, const odla_value_id value_id)
	Dequantize a tensor. More...
ODLA_API_EXPORT odla_value ODLA_API_CALL	odla_Quantize (odla_value input, odla_value scale, odla_value zero_point, odla_int32 axis, odla_element_type target_data_type, const odla_value_id value_id)
	Quantize a tensor. More...

Detailed Description

This file defines the ODLA quantization related operators.

Definition in file odla_ops_quantization.h.

Function Documentation

odla_Dequantize()

ODLA_API_EXPORT odla_value ODLA_API_CALL odla_Dequantize	(	odla_value	input,
		odla_value	scale,
		odla_value	zero_point,
		odla_int32	axis,
		odla_element_type	target_data_type,
		const odla_value_id	value_id
	)

Dequantize a tensor.

Converts a quantized tensor to the full precision tensor.

Parameters

input	the input value
scale	scale for input
zero_point	zero point for input
axis	the axis of the dequantizing. Ignored if scale is a scalar
target_data_type	the data type of full precision output
value_id	a unique value id (can be NULL)

Returns

odla_value

odla_Quantize()

ODLA_API_EXPORT odla_value ODLA_API_CALL odla_Quantize	(	odla_value	input,
		odla_value	scale,
		odla_value	zero_point,
		odla_int32	axis,
		odla_element_type	target_data_type,
		const odla_value_id	value_id
	)

Quantize a tensor.

Converts a tensor to low precision tensor.

Parameters

input	the input value
scale	scale for input
zero_point	zero point for input
axis	the axis of the quantizing. Ignored if scale is a scalar
target_data_type	the data type of low precision output
value_id	a unique value id (can be NULL)

Returns

odla_value