:py:mod:`neural_compressor.adaptor.tf_utils.smooth_quant_calibration`
=====================================================================

.. py:module:: neural_compressor.adaptor.tf_utils.smooth_quant_calibration

.. autoapi-nested-parse::

   Tensorflow model calibration process for Smooth Quantization.



Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   neural_compressor.adaptor.tf_utils.smooth_quant_calibration.SmoothQuantCalibration
   neural_compressor.adaptor.tf_utils.smooth_quant_calibration.SmoothQuantCalibrationLLM




.. py:class:: SmoothQuantCalibration(model, dataloader, iterations, op_types, percentile)


   A class for performing smooth quantization calibration on a Tensorflow model.

   :param model: The Tensorflow wrapper model to be calibrated.
   :type model: Model
   :param dataloader: The data loader for the calibration dataset.
   :type dataloader: DataLoader
   :param iterations: The number of iterations to run the calibration process.
   :type iterations: int
   :param op_types: The types of operations to be quantized.
   :type op_types: List[str]
   :param percentile: The percentile of calibration to remove outliers.
   :type percentile: float


.. py:class:: SmoothQuantCalibrationLLM(model_path, dataloader, iterations, op_types, percentile, temp_path, weight_name_mapping)




   A class for performing smooth quantization calibration on a Tensorflow LLM model.

   :param model: A path to the original Tensorflow model.
   :type model: str
   :param iterations: The number of iterations to run the calibration process.
   :type iterations: int
   :param op_types: The types of operations to be quantized.
   :type op_types: List[str]
   :param percentile: The percentile of calibration to remove outliers.
   :type percentile: float
   :param eval_func: The function to inference the model.
   :type eval_func: function
   :param temp_path: The temporary path to store median model.
   :type temp_path: str
   :param weight_name_mapping (): A function that convert weight tensor name in autotrackable to node name in graph_def