neural_compressor.tensorflow.quantization.quantize

Module Contents

Functions

quantize_model(model, quant_config[, ...])

The main entry to quantize model.

quantize_model_with_single_config(q_model, quant_config)

Quantize model using single config.
neural_compressor.tensorflow.quantization.quantize.quantize_model(model: str | tensorflow.keras.Model | neural_compressor.tensorflow.utils.BaseModel, quant_config: neural_compressor.common.base_config.BaseConfig | list, calib_dataloader: Callable = None, calib_iteration: int = 100)[source]

The main entry to quantize model.

Parameters:

  • model – a fp32 model to be quantized.

  • quant_config – single or lists of quantization configuration.

  • calib_dataloader – a data loader for calibration.

  • calib_iteration – the iteration of calibration.

Returns:

the quantized model.

Return type:

q_model

neural_compressor.tensorflow.quantization.quantize.quantize_model_with_single_config(q_model: neural_compressor.tensorflow.utils.BaseModel, quant_config: neural_compressor.common.base_config.BaseConfig, calib_dataloader: Callable = None, calib_iteration: int = 100)[source]

Quantize model using single config.

Parameters:

  • model – a model wrapped by INC TF model class.

  • quant_config – a quantization configuration.

  • calib_dataloader – a data loader for calibration.

  • calib_iteration – the iteration of calibration.

Returns:

the quantized model.

Return type:

q_model