sensitivity_anal.py 5.3 KB
Newer Older
L
LDOUBLEV 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
# Copyright (c) 2021 PaddlePaddle Authors. 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.

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import os
import sys

__dir__ = os.path.dirname(__file__)
sys.path.append(__dir__)
sys.path.append(os.path.join(__dir__, '..', '..', '..'))
sys.path.append(os.path.join(__dir__, '..', '..', '..', 'tools'))

L
LDOUBLEV 已提交
27 28 29 30 31 32 33 34
import json
import cv2
import paddle
from paddle import fluid
import paddleslim as slim
from copy import deepcopy
from tools import program

L
LDOUBLEV 已提交
35 36 37 38 39 40 41 42 43 44 45 46 47 48
import paddle
import paddle.distributed as dist
from ppocr.data import build_dataloader
from ppocr.modeling.architectures import build_model
from ppocr.losses import build_loss
from ppocr.optimizer import build_optimizer
from ppocr.postprocess import build_post_process
from ppocr.metrics import build_metric
from ppocr.utils.save_load import init_model
import tools.program as program

dist.get_world_size()


L
LDOUBLEV 已提交
49 50 51 52 53 54 55 56 57 58 59 60 61 62
def get_pruned_params(parameters, mode="det"):
    if mode == "det":
        skip_prune_params = [
            "conv2d_56.w_0", "conv2d_54.w_0", "conv2d_51.w_0",
            "conv_last_weights", "conv14_linear_weights",
            "conv13_expand_weights", "conv12_linear_weights",
            "conv12_expand_weights", "conv7_expand_weights",
            "conv8_expand_weights", "conv8_linear_weights",
            "conv5_linear_weights", "conv5_expand_weights",
            "conv3_linear_weights"
        ]
        skip_prune_params = skip_prune_params + ['conv2d_53.w_0']
    else:
        skip_prune_params = None
L
LDOUBLEV 已提交
63 64 65 66 67 68
    params = []

    for param in parameters:
        if len(
                param.shape
        ) == 4 and 'depthwise' not in param.name and 'transpose' not in param.name and "conv2d_57" not in param.name and "conv2d_56" not in param.name:
L
LDOUBLEV 已提交
69 70
            if param.name not in skip_prune_params:
                params.append(param.name)
L
LDOUBLEV 已提交
71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
    return params


def main(config, device, logger, vdl_writer):
    # init dist environment
    if config['Global']['distributed']:
        dist.init_parallel_env()

    global_config = config['Global']

    # build dataloader
    train_dataloader = build_dataloader(config, 'Train', device, logger)
    if config['Eval']:
        valid_dataloader = build_dataloader(config, 'Eval', device, logger)
    else:
        valid_dataloader = None

    # build post process
    post_process_class = build_post_process(config['PostProcess'],
                                            global_config)

    # build model
    # for rec algorithm
    if hasattr(post_process_class, 'character'):
        char_num = len(getattr(post_process_class, 'character'))
        config['Architecture']["Head"]['out_channels'] = char_num
    model = build_model(config['Architecture'])

    flops = paddle.flops(model, [1, 3, 640, 640])
L
LDOUBLEV 已提交
100
    print(f"FLOPs before pruning: {flops}")
L
LDOUBLEV 已提交
101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129

    from paddleslim.dygraph import FPGMFilterPruner
    model.train()
    pruner = FPGMFilterPruner(model, [1, 3, 640, 640])

    # build loss
    loss_class = build_loss(config['Loss'])

    # build optim
    optimizer, lr_scheduler = build_optimizer(
        config['Optimizer'],
        epochs=config['Global']['epoch_num'],
        step_each_epoch=len(train_dataloader),
        parameters=model.parameters())

    # build metric
    eval_class = build_metric(config['Metric'])
    # load pretrain model
    pre_best_model_dict = init_model(config, model, logger, optimizer)

    logger.info('train dataloader has {} iters, valid dataloader has {} iters'.
                format(len(train_dataloader), len(valid_dataloader)))

    def eval_fn():
        metric = program.eval(model, valid_dataloader, post_process_class,
                              eval_class)
        logger.info(f"metric['hmean']: {metric['hmean']}")
        return metric['hmean']

L
LDOUBLEV 已提交
130
    pruner.sensitive(
L
LDOUBLEV 已提交
131 132 133 134 135 136
        eval_func=eval_fn,
        sen_file="./sen.pickle",
        skip_vars=[
            "conv2d_57.w_0", "conv2d_transpose_2.w_0", "conv2d_transpose_3.w_0"
        ])

L
LDOUBLEV 已提交
137 138 139 140 141
    params = get_pruned_params(model.parameters())
    ratios = {}
    # set the prune ratio is 0.2
    for param in params:
        ratios[param] = 0.2
L
LDOUBLEV 已提交
142

L
LDOUBLEV 已提交
143
    plan = pruner.prune_vars(ratios, [0])
L
LDOUBLEV 已提交
144 145 146
    for param in model.parameters():
        if ("weights" in param.name and "conv" in param.name) or (
                "w_0" in param.name and "conv2d" in param.name):
L
LDOUBLEV 已提交
147
            print(f"{param.name}: {param.shape}")
L
LDOUBLEV 已提交
148 149

    flops = paddle.flops(model, [1, 3, 640, 640])
L
LDOUBLEV 已提交
150
    print(f"FLOPs after pruning: {flops}")
L
LDOUBLEV 已提交
151 152 153 154 155 156 157 158 159 160

    # start train
    program.train(config, train_dataloader, valid_dataloader, device, model,
                  loss_class, optimizer, lr_scheduler, post_process_class,
                  eval_class, pre_best_model_dict, logger, vdl_writer)


if __name__ == '__main__':
    config, device, logger, vdl_writer = program.preprocess(is_train=True)
    main(config, device, logger, vdl_writer)