hard_swish

Classes

HardSwishOperator

HardSwishOperator(op: AirOperator, model: AirModel, platform: SocPlatform, prefix: str = 'aot', attributes: dict[str, str] = {})

HARD_SWISH operator.

This operator computes the hard swish activation function on the input tensor.

Parameters:

• op

  (AirOperator) –

  The AIR operator to wrap.

• model

  (AirModel) –

  The AIR model.

• platform

  (SocPlatform) –

  The target platform for code generation.

• prefix

  (str, default: 'aot' ) –

  Prefix for generated code files. Defaults to "aot".

• attributes

  (dict[str, str], default: {} ) –

  Attributes for template values. Defaults to {}.

Functions

compute_values

compute_values() -> dict[str, str]

Compute the values for the operator.

Returns:

• dict[str, str] –

  dict[str, str]: Computed values for the operator.

emit

emit(save_path: Path)

Generate the source code for the operator.

Parameters:

• save_path

  (Path) –

  Path to save the generated code.

Functions

compute_prescale

compute_prescale(q6: int, dtype: dtype) -> int

Compute the prescale shift needed to keep intermediate products in range.

Parameters:

• q6

  (int) –

  The quantized value representing 6.0 in the input scale

• dtype

  (dtype) –

  The data type of the input tensor (np.int8 or np.int16)

Returns:

• int ( int ) –

  The prescale shift amount

make_hardswish_lut_s8

make_hardswish_lut_s8(input_scale: float, input_zero_point: int, output_scale: float, output_zero_point: int) -> np.ndarray

Build a 256-entry int8 → int8 LUT for HardSwish, iterating over q in [-128,127].

Parameters:

• input_scale

  (float) –

  Scale of the input tensor.

• input_zero_point

  (int) –

  Zero point of the input tensor.

• output_scale

  (float) –

  Scale of the output tensor.

• output_zero_point

  (int) –

  Zero point of the output tensor.

Returns:

• ndarray –

  np.ndarray: The generated LUT.