neural_compressor.pruner.pruning

Pruning.

Module Contents

Classes

Pruning

Pruning.

class neural_compressor.pruner.pruning.Pruning(config)

Pruning.

The main class to do pruning; it contains at least one Pruner object.

Parameters:

config – a string representing the path to a config file. For config file template, please refer to https://github.com/intel/neural-compressor/tree/master/examples/pytorch/nlp/huggingface_models/text-classification/pruning/pytorch_pruner/eager/

model

The model object to prune.

config_file_path

A string representing the path to a config file.

pruners

A list. A list of Pruner objects.

pruner_info

A config dict object that contains pruners’ information.

property model

Obtain model in neural_compressor.model.

update_config(*args, **kwargs)

Add user-defined arguments to the original configurations.

The original config of pruning is read from a file. However, users can still modify configurations by passing key-value arguments in this function. Please note that the key-value arguments’ keys could be processed in current configuration.

get_sparsity_ratio()

Calculate sparsity ratio of a module/layer.

Returns:

Three floats. elementwise_over_matmul_gemm_conv refers to zero elements’ ratio in pruning layers. elementwise_over_all refers to zero elements’ ratio in all layers in the model. blockwise_over_matmul_gemm_conv refers to all-zero blocks’ ratio in pruning layers.

on_train_begin()

Implement at the beginning of training process.

Before training, ensure that pruners are generated.

on_epoch_begin(epoch)

Implement at the beginning of every epoch.

on_step_begin(local_step)

Implement at the beginning of every step.

on_before_optimizer_step()

Implement before optimizer.step().

on_step_end()

Implement at the end of every step.

on_epoch_end()

Implement the end of every epoch.

on_train_end()

Implement the end of training phase.

on_before_eval()

Implement at the beginning of evaluation phase.

on_after_eval()

Implement at the end of evaluation phase.

on_after_optimizer_step()

Implement after optimizer.step().