Dubhe/model_measuring/kamal/slim/prunning/strategy.py

"""
 Copyright 2020 Tianshu AI Platform. All Rights Reserved.

 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
 You may obtain a copy of the License at

     http://www.apache.org/licenses/LICENSE-2.0

 Unless required by applicable law or agreed to in writing, software
 distributed under the License is distributed on an "AS IS" BASIS,
 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License.
 =============================================================
"""

import torch_pruning as tp
import abc
import torch
import torch.nn as nn
import random
import numpy as np 

_PRUNABLE_MODULES= tp.DependencyGraph.PRUNABLE_MODULES

class BaseStrategy(abc.ABC):

    @abc.abstractmethod
    def select(self, layer_to_prune):
        pass

    def  __call__(self, model, rate=0.1, example_inputs=None):
        if example_inputs is None:
            example_inputs = torch.randn( 1,3,256,256 )

        DG = tp.DependencyGraph()
        DG.build_dependency(model, example_inputs=example_inputs)

        prunable_layers = []
        total_params = 0
        num_accumulative_conv_params = [ 0, ]

        for m in model.modules():
            if isinstance(m, _PRUNABLE_MODULES ) :
                nparam = tp.utils.count_prunable_params( m )
                total_params += nparam
                if isinstance(m, (nn.modules.conv._ConvNd, nn.Linear)):
                    prunable_layers.append( m )
                    num_accumulative_conv_params.append( num_accumulative_conv_params[-1]+nparam )
        prunable_layers.pop(-1) # remove the last layer
        num_accumulative_conv_params.pop(-1) # remove the last layer
        
        num_conv_params = num_accumulative_conv_params[-1]
        num_accumulative_conv_params = [ ( num_accumulative_conv_params[i], num_accumulative_conv_params[i+1] ) for i in range(len(num_accumulative_conv_params)-1) ]
        
        def map_param_idx_to_conv_layer(i):
            for l, accu in zip( prunable_layers, num_accumulative_conv_params ):
                if accu[0]<=i and i<accu[1]:
                    return l 

        num_pruned = 0
        while num_pruned<total_params*rate:
            layer_to_prune = map_param_idx_to_conv_layer( random.randint( 0, num_conv_params-1 ) )
            if layer_to_prune.weight.shape[0]<1:
                continue
            idx = self.select( layer_to_prune )
            fn = tp.prune_conv if isinstance(layer_to_prune, nn.modules.conv._ConvNd) else tp.prune_linear
            plan = DG.get_pruning_plan( layer_to_prune, fn, idxs=idx ) 
            num_pruned += plan.exec() 
        return model

class RandomStrategy(BaseStrategy):
    def select(self, layer_to_prune):
        return [ random.randint( 0, layer_to_prune.weight.shape[0]-1 ) ]

class LNStrategy(BaseStrategy):
    def __init__(self, n=2):
        self.n = n

    def select(self, layer_to_prune):
        w = torch.flatten( layer_to_prune.weight, 1 )
        norm = torch.norm(w, p=self.n, dim=1)
        idx = [ int(norm.min(dim=0)[1].item()) ]
        return idx