:py:mod:`neural_compressor.compression.pruner.model_slim.auto_slim`
===================================================================

.. py:module:: neural_compressor.compression.pruner.model_slim.auto_slim

.. autoapi-nested-parse::

   Auto slim.



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


Functions
~~~~~~~~~

.. autoapisummary::

   neural_compressor.compression.pruner.model_slim.auto_slim.model_slim
   neural_compressor.compression.pruner.model_slim.auto_slim.model_slim_ffn2
   neural_compressor.compression.pruner.model_slim.auto_slim.model_slim_mha
   neural_compressor.compression.pruner.model_slim.auto_slim.parse_auto_slim_config
   neural_compressor.compression.pruner.model_slim.auto_slim.generate_ffn2_pruning_config
   neural_compressor.compression.pruner.model_slim.auto_slim.generate_mha_pruning_config



.. py:function:: model_slim(model, dataloader=None, round_multiplier=32)

   Slim the sparse model automatically.


.. py:function:: model_slim_ffn2(model, dataloader=None, round_multiplier=32)

   Remove some sparse part in the model permanently and obtain acceleration directly.

   :param model: a sprase model.
   :param round_multiplier: the channel number after slimming should be multiple of this number.
   :type round_multiplier: int


.. py:function:: model_slim_mha(model, dataloader=None)

   Remove some sparse part in the model permanently and obtain acceleration directly.

   :param model: a sprase model.


.. py:function:: parse_auto_slim_config(model, dataloader=None, ffn2_sparsity=0.0, mha_sparsity=0.0, **kwargs)

   Get model slim pruning configs.


.. py:function:: generate_ffn2_pruning_config(model, dataloader, ffn2_sparsity, **kwargs)

   Get consecutive linear layers pruning configs.


.. py:function:: generate_mha_pruning_config(model, dataloader, mha_sparsity, **kwargs)

   Get multi-head attention layers pruning configs.