diff --git a/slim/sensitive/README.md b/slim/sensitive/README.md new file mode 100644 index 0000000000000000000000000000000000000000..ae0bba0a9910937830ffae9145319a6be02223e7 --- /dev/null +++ b/slim/sensitive/README.md @@ -0,0 +1,96 @@ +# 卷积层敏感度分析教程 + +请确保已正确[安装PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.1/docs/INSTALL_cn.md)及其依赖。 + +该文档介绍如何使用[PaddleSlim](https://paddlepaddle.github.io/PaddleSlim)的敏感度分析接口对检测库中的模型的卷积层进行敏感度分析。 + +在检测库中，可以直接调用`PaddleDetection/slim/sensitive/sensitive.py`脚本实现敏感度分析，在该脚本中调用了PaddleSlim的[paddleslim.prune.sensitivity](https://paddlepaddle.github.io/PaddleSlim/api/prune_api/#sensitivity)接口。 + +该教程中所示操作，如无特殊说明，均在`PaddleDetection/slim/sensitive/`路径下执行。 + +## 数据准备 + +请参考检测库[数据模块](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.1/docs/DATA_cn.md)文档准备数据。 + +## 模型选择 + +通过`-c`选项指定待分析模型的配置文件的相对路径，更多可选配置文件请参考: [检测库配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/release/0.1/configs) + +通过`-o weights`指定模型的权重，可以指定url或本地文件系统的路径。如下所示： + +``` +-o weights=https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar +``` + +或 + +``` +-o weights=output/yolov3_mobilenet_v1_voc/model_final +``` + +官方已发布的模型请参考: [模型库](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.1/docs/MODEL_ZOO_cn.md) + +## 确定待分析参数 + +在计算敏感度之前，需要查出待分析的卷积层的参数的名称。通过以下命令查看当前模型的所有参数： + +``` +python sensitive.py \ +-c ../../configs/yolov3_mobilenet_v1_voc.yml \ +--print_params +``` + +通过观察参数名称和参数的形状，筛选出所有卷积层参数，并确定要分析的卷积层参数。 + +## 执行分析 + +通过选项`--pruned_params`指定待分析的卷积层参数名，参数名间以英文字符逗号分割。 +通过选项`--sensitivities_file`指定敏感度信息保存的文件，敏感度信息会追加到该文件中。重启敏感度计算任务，该文件中已计算的信息不会再被计算。 + +示例如下： + +``` +nohup python sensitive.py \ +-c ../../configs/yolov3_mobilenet_v1_voc.yml \ +--pruned_params "yolo_block.0.0.0.conv.weights,yolo_block.0.0.1.conv.weights,yolo_block.0.1.0.conv.weights,yolo_block.0.1.1.conv.weights,yolo_block.0.2.conv.weights,yolo_block.0.tip.conv.weights,yolo_block.1.0.0.conv.weights,yolo_block.1.0.1.conv.weights,yolo_block.1.1.0.conv.weights,yolo_block.1.1.1.conv.weights,yolo_block.1.2.conv.weights,yolo_block.1.tip.conv.weights,yolo_block.2.0.0.conv.weights,yolo_block.2.0.1.conv.weights,yolo_block.2.1.0.conv.weights,yolo_block.2.1.1.conv.weights,yolo_block.2.2.conv.weights,yolo_block.2.tip.conv.weights" \ +--sensitivities_file "./demo.data" +``` + +执行`python sensitive.py --help`查看更多选项。 + +## 分析敏感度信息 + +可以通过[paddleslim.prune.load_sensitivities](https://paddlepaddle.github.io/PaddleSlim/api/prune_api/#load_sensitivities)从文件中加载敏感度信息，并使用Python数据分析工具画图分析。下图展示了`MobileNetv1-YOLOv3-VOC`模型在VOC数据上的敏感度信息： + +
+ +
+ +通过画图分析，可以确定一组合适的剪裁率，或者通过[paddleslim.prune.get_ratios_by_loss](https://paddlepaddle.github.io/PaddleSlim/api/prune_api/#get_ratios_by_losssensitivities-loss)获得合适的剪裁率。 + +## 分布式计算敏感度信息 + +如果模型评估速度比较慢，可以考虑使用多进程加速敏感度计算的过程。 +通过`--pruned_ratios`指定当前进程计算敏感度时用的剪裁率，默认为"0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9"。可以将该组剪切率分配到不同的进程进行计算，如下所示： + +``` +# 进程1 +nohup python sensitive.py \ +-c ../../configs/yolov3_mobilenet_v1_voc.yml \ +--pruned_params "yolo_block.0.0.0.conv.weights" \ +--pruned_ratios "0.1 0.2 0.3 0.4 0.5" +--sensitivities_file "./demo.data.1" +``` + +``` +# 进程2 +nohup python sensitive.py \ +-c ../../configs/yolov3_mobilenet_v1_voc.yml \ +--pruned_params "yolo_block.0.0.0.conv.weights" \ +--pruned_ratios "0.6 0.7 0.8 0.9" +--sensitivities_file "./demo.data.2" +``` + +待以上两个进程执行完毕，通过[paddleslim.prune.merge_sensitive](https://paddlepaddle.github.io/PaddleSlim/api/prune_api/#merge_sensitive)将`demo.data.1`和`demo.data.2`两个文件合并分析。 + + diff --git a/slim/sensitive/images/mobilev1_yolov3_voc_sensitives.png b/slim/sensitive/images/mobilev1_yolov3_voc_sensitives.png new file mode 100644 index 0000000000000000000000000000000000000000..a9609e89425d28164423cb5e3281206c6be6969f Binary files /dev/null and b/slim/sensitive/images/mobilev1_yolov3_voc_sensitives.png differ diff --git a/slim/sensitive/sensitive.py b/slim/sensitive/sensitive.py new file mode 100644 index 0000000000000000000000000000000000000000..2e8f41c58ce0bff0e1d852f5057eaace1ae8819f --- /dev/null +++ b/slim/sensitive/sensitive.py @@ -0,0 +1,206 @@ +# Copyright (c) 2019 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 time +import numpy as np +import datetime +from collections import deque + + +def set_paddle_flags(**kwargs): + for key, value in kwargs.items(): + if os.environ.get(key, None) is None: + os.environ[key] = str(value) + + +# NOTE(paddle-dev): All of these flags should be set before +# `import paddle`. Otherwise, it would not take any effect. +set_paddle_flags( + FLAGS_eager_delete_tensor_gb=0, # enable GC to save memory +) + + +from paddle import fluid +from ppdet.experimental import mixed_precision_context +from ppdet.core.workspace import load_config, merge_config, create +#from ppdet.data.data_feed import create_reader + +from ppdet.data.reader import create_reader + +from ppdet.utils.cli import print_total_cfg +from ppdet.utils import dist_utils +from ppdet.utils.eval_utils import parse_fetches, eval_run, eval_results +from ppdet.utils.stats import TrainingStats +from ppdet.utils.cli import ArgsParser +from ppdet.utils.check import check_gpu, check_version +import ppdet.utils.checkpoint as checkpoint +from ppdet.modeling.model_input import create_feed +from paddleslim.prune import sensitivity +import logging +FORMAT = '%(asctime)s-%(levelname)s: %(message)s' +logging.basicConfig(level=logging.INFO, format=FORMAT) +logger = logging.getLogger(__name__) + + +def main(): + env = os.environ + + print("FLAGS.config: {}".format(FLAGS.config)) + cfg = load_config(FLAGS.config) + assert 'architecture' in cfg + main_arch = cfg.architecture + + merge_config(FLAGS.opt) + + print_total_cfg(cfg) + + place = fluid.CUDAPlace(0) + exe = fluid.Executor(place) + + # build program + startup_prog = fluid.Program() + eval_prog = fluid.Program() + with fluid.program_guard(eval_prog, startup_prog): + with fluid.unique_name.guard(): + model = create(main_arch) + inputs_def = cfg['EvalReader']['inputs_def'] + feed_vars, eval_loader = model.build_inputs(**inputs_def) + fetches = model.eval(feed_vars) + eval_prog = eval_prog.clone(True) + + if FLAGS.print_params: + print("-------------------------All parameters in current graph----------------------") + for block in eval_prog.blocks: + for param in block.all_parameters(): + print("parameter name: {}\tshape: {}".format(param.name, param.shape)) + print("------------------------------------------------------------------------------") + return + + eval_reader = create_reader(cfg.EvalReader) + eval_loader.set_sample_list_generator(eval_reader, place) + + # parse eval fetches + extra_keys = [] + if cfg.metric == 'COCO': + extra_keys = ['im_info', 'im_id', 'im_shape'] + if cfg.metric == 'VOC': + extra_keys = ['gt_box', 'gt_label', 'is_difficult'] + if cfg.metric == 'WIDERFACE': + extra_keys = ['im_id', 'im_shape', 'gt_box'] + eval_keys, eval_values, eval_cls = parse_fetches(fetches, eval_prog, + extra_keys) + + exe.run(startup_prog) + + fuse_bn = getattr(model.backbone, 'norm_type', None) == 'affine_channel' + + ignore_params = cfg.finetune_exclude_pretrained_params \ + if 'finetune_exclude_pretrained_params' in cfg else [] + + start_iter = 0 + + if cfg.weights: + checkpoint.load_params(exe, eval_prog, cfg.weights) + else: + logger.warn("Please set cfg.weights to load trained model.") + + # whether output bbox is normalized in model output layer + is_bbox_normalized = False + if hasattr(model, 'is_bbox_normalized') and \ + callable(model.is_bbox_normalized): + is_bbox_normalized = model.is_bbox_normalized() + + # if map_type not set, use default 11point, only use in VOC eval + map_type = cfg.map_type if 'map_type' in cfg else '11point' + + def test(program): + + compiled_eval_prog = fluid.compiler.CompiledProgram(program) + + results = eval_run(exe, compiled_eval_prog, eval_loader, + eval_keys, eval_values, eval_cls) + resolution = None + if 'mask' in results[0]: + resolution = model.mask_head.resolution + dataset = cfg['EvalReader']['dataset'] + box_ap_stats = eval_results( + results, + cfg.metric, + cfg.num_classes, + resolution, + is_bbox_normalized, + FLAGS.output_eval, + map_type, + dataset=dataset) + return box_ap_stats[0] + + pruned_params = FLAGS.pruned_params + + assert (FLAGS.pruned_params is not None), "FLAGS.pruned_params is empty!!! Please set it by '--pruned_params' option." + pruned_params = FLAGS.pruned_params.strip().split(",") + logger.info("pruned params: {}".format(pruned_params)) + pruned_ratios = [float(n) for n in FLAGS.pruned_ratios.strip().split(" ")] + logger.info("pruned ratios: {}".format(pruned_ratios)) + sensitivity(eval_prog, + place, + pruned_params, + test, + sensitivities_file=FLAGS.sensitivities_file, + pruned_ratios=pruned_ratios) + + +if __name__ == '__main__': + parser = ArgsParser() + parser.add_argument( + "--output_eval", + default=None, + type=str, + help="Evaluation directory, default is current directory.") + parser.add_argument( + "-d", + "--dataset_dir", + default=None, + type=str, + help="Dataset path, same as DataFeed.dataset.dataset_dir") + parser.add_argument( + "-s", + "--sensitivities_file", + default="sensitivities.data", + type=str, + help="The file used to save sensitivities.") + parser.add_argument( + "-p", + "--pruned_params", + default=None, + type=str, + help="The parameters to be pruned when calculating sensitivities.") + parser.add_argument( + "-r", + "--pruned_ratios", + default="0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9", + type=str, + help="The ratios pruned iteratively for each parameter when calculating sensitivities.") + parser.add_argument( + "-P", + "--print_params", + default=False, + action='store_true', + help="Whether to only print the parameters' names and shapes.") + FLAGS = parser.parse_args() + main()