utils

Classes

Functions

quantize_multiplier

quantize_multiplier(real_multiplier: float) -> tuple[int, int]

Convert a floating point multiplier into a fixed-point multiplier and shift.

Given a real multiplier, this function computes a pair (quantized_multiplier, shift) such that the fixed-point multiplication approximates the real multiplier:

real_multiplier ~ quantized_multiplier / (2^(31 - shift))

This is typically used in quantized inference to convert floating-point scaling factors to fixed-point representation.

Parameters:

• real_multiplier

  (float) –

  The floating-point multiplier.

Returns:

• tuple ( tuple[int, int] ) –

  (quantized_multiplier, shift)

compute_fused_activation_range

compute_fused_activation_range(activation_function: AirActivationType, output_zero_point: int = 0, output_scale: float = 1.0, qmin: int = -128, qmax: int = 127) -> tuple[int, int]

Compute the activation min and max values based on the fused activation function.

Parameters:

• activation_function

  (ActivationFunctionType) –

  The activation function type.

• output_zero_point

  (int, default: 0 ) –

  The zero point for the output tensor. Defaults to 0.

• output_scale

  (float, default: 1.0 ) –

  The scale for the output tensor. Defaults to 1.0.

• qmin

  (int, default: -128 ) –

  Minimum quantized value. Defaults to -128.

• qmax

  (int, default: 127 ) –

  Maximum quantized value. Defaults to 127.

Returns:

• tuple[int, int] –

  tuple[int, int]: The activation min and max values.

compute_padding_with_offset

compute_padding_with_offset(in_size: int, filter_size: int, stride: int, dilation_rate: int = 1, padding_type: int = AirPaddingType.SAME) -> tuple[int, int]

Computes a single padding value (top/left) for one dimension with offset.

Parameters:

• in_size

  (int) –

  Input dimension (height or width).

• filter_size

  (int) –

  Kernel dimension (height or width).

• stride

  (int) –

  Stride for that dimension.

• dilation_rate

  (int, default: 1 ) –

  Dilation rate for that dimension (default is 1).

• padding_type

  (int, default: SAME ) –

  0 for SAME padding, 1 for VALID padding (default is SAME).

Returns:

• int ( tuple[int, int] ) –

  The computed top or left padding value.

compute_padding

compute_padding(in_size: int, filter_size: int, stride: int, dilation_rate: int = 1, padding_type: int = AirPaddingType.SAME) -> int

Computes a single padding value (top/left) for one dimension.

Parameters:

• in_size

  (int) –

  Input dimension (height or width).

• filter_size

  (int) –

  Kernel dimension (height or width).

• stride

  (int) –

  Stride for that dimension.

• dilation_rate

  (int, default: 1 ) –

  Dilation rate for that dimension (default is 1).

• padding_type

  (int, default: SAME ) –

  0 for SAME padding, 1 for VALID padding (default is SAME).

Returns:

• int ( int ) –

  The computed top or left padding value.

get_expanded_dims_shape

get_expanded_dims_shape(x: NDArray, num_dims: int = 4, pre: bool = True) -> tuple[int, ...]

Expand the shape of a tensor to a specified number of dimensions.

Parameters:

• x

  (NDArray) –

  Input tensor whose shape is to be expanded.

• num_dims

  (int, default: 4 ) –

  Number of dimensions to expand to. Defaults to 4.

Returns:

• tuple[int, ...] –

  tuple[int, ...]: Expanded shape of the tensor.

reduce_multiplier_q31_to_q15

reduce_multiplier_q31_to_q15(multiplier: int) -> int

Reduce a multiplier value for int16 quantization.

Certain CMSIS functions for int16 quantized models require the multiplier to be reduced from Q31 format to Q15 format. This function performs that reduction.

Parameters:

• multiplier

  (int) –

  The original multiplier value.

Returns:

• int ( int ) –

  The reduced multiplier value.