From 1a843c55b457213b6f7033ffc972411557cbb956 Mon Sep 17 00:00:00 2001 From: Bai Yifan Date: Fri, 3 Jan 2020 10:22:38 +0800 Subject: [PATCH] Update distillation demo (#128) * update distillation demo --- slim/distillation/README.md | 183 +++++----- slim/distillation/compress.py | 324 ------------------ slim/distillation/distill.py | 289 ++++++++++++++++ slim/distillation/run.sh | 47 --- ...ilenet_v1_yolov3_resnet34_distillation.yml | 18 - slim/distillation/yolov3_resnet34.yml | 34 -- 6 files changed, 391 insertions(+), 504 deletions(-) mode change 100755 => 100644 slim/distillation/README.md delete mode 100644 slim/distillation/compress.py create mode 100644 slim/distillation/distill.py delete mode 100644 slim/distillation/run.sh delete mode 100644 slim/distillation/yolov3_mobilenet_v1_yolov3_resnet34_distillation.yml delete mode 100644 slim/distillation/yolov3_resnet34.yml diff --git a/slim/distillation/README.md b/slim/distillation/README.md old mode 100755 new mode 100644 index e2666cd01..dee4aaafe --- a/slim/distillation/README.md +++ b/slim/distillation/README.md @@ -1,141 +1,162 @@ ->运行该示例前请安装Paddle1.6或更高版本 +>运行该示例前请安装PaddleSlim和Paddle1.6或更高版本 # 检测模型蒸馏示例 ## 概述 -该示例使用PaddleSlim提供的[蒸馏策略](https://github.com/PaddlePaddle/models/blob/develop/PaddleSlim/docs/tutorial.md#3-蒸馏)对检测库中的模型进行蒸馏训练。 +该示例使用PaddleSlim提供的[蒸馏策略](https://paddlepaddle.github.io/PaddleSlim/algo/algo/#3)对检测库中的模型进行蒸馏训练。 在阅读该示例前，建议您先了解以下内容： - [检测库的常规训练方法](https://github.com/PaddlePaddle/PaddleDetection) -- [PaddleSlim使用文档](https://github.com/PaddlePaddle/models/blob/develop/PaddleSlim/docs/usage.md) +- [PaddleSlim蒸馏API文档](https://paddlepaddle.github.io/PaddleSlim/api/single_distiller_api/) +## 安装PaddleSlim +可按照[PaddleSlim使用文档](https://paddlepaddle.github.io/PaddleSlim/)中的步骤安装PaddleSlim -## 配置文件说明 +## 蒸馏策略说明 -关于配置文件如何编写您可以参考： +关于蒸馏API如何使用您可以参考PaddleSlim蒸馏API文档 -- [PaddleSlim配置文件编写说明](https://github.com/PaddlePaddle/models/blob/develop/PaddleSlim/docs/usage.md#122-%E9%85%8D%E7%BD%AE%E6%96%87%E4%BB%B6%E7%9A%84%E4%BD%BF%E7%94%A8) -- [蒸馏策略配置文件编写说明](https://github.com/PaddlePaddle/models/blob/develop/PaddleSlim/docs/usage.md#23-蒸馏) - -这里以ResNet34-YoloV3蒸馏MobileNetV1-YoloV3模型为例，首先，为了对`student model`和`teacher model`有个总体的认识，从而进一步确认蒸馏的对象，我们通过以下命令分别观察两个网络变量（Variable）的名称和形状： +这里以ResNet34-YoloV3蒸馏训练MobileNetV1-YoloV3模型为例，首先，为了对`student model`和`teacher model`有个总体的认识，进一步确认蒸馏的对象，我们通过以下命令分别观察两个网络变量（Variables）的名称和形状： ```python -# 观察student model的Variable +# 观察student model的Variables +student_vars = [] for v in fluid.default_main_program().list_vars(): - if "py_reader" not in v.name and "double_buffer" not in v.name and "generated_var" not in v.name: - print(v.name, v.shape) -# 观察teacher model的Variable + try: + student_vars.append((v.name, v.shape)) + except: + pass +print("="*50+"student_model_vars"+"="*50) +print(student_vars) +# 观察teacher model的Variables +teacher_vars = [] for v in teacher_program.list_vars(): - print(v.name, v.shape) + try: + teacher_vars.append((v.name, v.shape)) + except: + pass +print("="*50+"teacher_model_vars"+"="*50) +print(teacher_vars) ``` -经过对比可以发现，`student model`和`teacher model`的部分中间结果分别为： +经过对比可以发现，`student model`和`teacher model`输入到3个`yolov3_loss`的特征图分别为： ```bash # student model -conv2d_15.tmp_0 +conv2d_20.tmp_1, conv2d_28.tmp_1, conv2d_36.tmp_1 # teacher model -teacher_teacher_conv2d_1.tmp_0 +conv2d_6.tmp_1, conv2d_14.tmp_1, conv2d_22.tmp_1 ``` -所以，我们用`l2_distiller`对这两个特征图做蒸馏。在配置文件中进行如下配置： - -```yaml -distillers: - l2_distiller: - class: 'L2Distiller' - teacher_feature_map: 'teacher_teacher_conv2d_1.tmp_0' - student_feature_map: 'conv2d_15.tmp_0' - distillation_loss_weight: 1 -strategies: - distillation_strategy: - class: 'DistillationStrategy' - distillers: ['l2_distiller'] - start_epoch: 0 - end_epoch: 270 +它们形状两两相同，且分别处于两个网络的输出部分。所以，我们用`l2_loss`对这几个特征图两两对应添加蒸馏loss。需要注意的是，teacher的Variable在merge过程中被自动添加了一个`name_prefix`，所以这里也需要加上这个前缀`"teacher_"`，merge过程请参考[蒸馏API文档](https://paddlepaddle.github.io/PaddleSlim/api/single_distiller_api/#merge) + +```python +dist_loss_1 = l2_loss('teacher_conv2d_6.tmp_1', 'conv2d_20.tmp_1') +dist_loss_2 = l2_loss('teacher_conv2d_14.tmp_1', 'conv2d_28.tmp_1') +dist_loss_3 = l2_loss('teacher_conv2d_22.tmp_1', 'conv2d_36.tmp_1') ``` -我们也可以根据上述操作为蒸馏策略选择其他loss，PaddleSlim支持的有`FSP_loss`, `L2_loss`和`softmax_with_cross_entropy_loss` 。 +我们也可以根据上述操作为蒸馏策略选择其他loss，PaddleSlim支持的有`FSP_loss`, `L2_loss`, `softmax_with_cross_entropy_loss` 以及自定义的任何loss。 ## 训练 -根据[PaddleDetection/tools/train.py](https://github.com/PaddlePaddle/PaddleDetection/tree/master/tools/train.py)编写压缩脚本compress.py。 -在该脚本中定义了Compressor对象，用于执行压缩任务。 +根据[PaddleDetection/tools/train.py](../../tools/train.py)编写压缩脚本`distill.py`。 +在该脚本中定义了teacher_model和student_model，用teacher_model的输出指导student_model的训练 +### 执行示例 +step1: 设置GPU卡 +```shell +export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +``` -您可以通过运行脚本`run.sh`运行该示例。 +step2: 开始训练 +```bash +python slim/distillation/distill.py \ + -c configs/yolov3_mobilenet_v1_voc.yml \ + -t configs/yolov3_r34_voc.yml \ + --teacher_pretrained https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar +``` -### 保存断点（checkpoint） +如果要调整训练卡数，需要调整配置文件`yolov3_mobilenet_v1_voc.yml`中的以下参数： -如果在配置文件中设置了`checkpoint_path`, 则在蒸馏任务执行过程中会自动保存断点，当任务异常中断时， -重启任务会自动从`checkpoint_path`路径下按数字顺序加载最新的checkpoint文件。如果不想让重启的任务从断点恢复， -需要修改配置文件中的`checkpoint_path`，或者将`checkpoint_path`路径下文件清空。 +- **max_iters:** 训练过程迭代总步数。 +- **YOLOv3Loss.batch_size:** 该参数表示单张GPU卡上的`batch_size`, 总`batch_size`是GPU卡数乘以这个值， `batch_size`的设定受限于显存大小。 +- **LeaningRate.base_lr:** 根据多卡的总`batch_size`调整`base_lr`，两者大小正相关，可以简单的按比例进行调整。 +- **LearningRate.schedulers.PiecewiseDecay.milestones：** 请根据batch size的变化对其调整。 +- **LearningRate.schedulers.PiecewiseDecay.LinearWarmup.steps：** 请根据batch size的变化对其进行调整。 ->注意：配置文件中的信息不会保存在断点中，重启前对配置文件的修改将会生效。 +以下为4卡训练示例，通过命令行覆盖`yolov3_mobilenet_v1_voc.yml`中的参数： +```shell +CUDA_VISIBLE_DEVICES=0,1,2,3 +python slim/distillation/distill.py \ + -c configs/yolov3_mobilenet_v1_voc.yml \ + -t configs/yolov3_r34_voc.yml \ + -o YoloTrainFeed.batch_size=16 \ + --teacher_pretrained https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar +``` -## 评估 -如果在配置文件中设置了`checkpoint_path`，则每个epoch会保存一个压缩后的用于评估的模型， -该模型会保存在`${checkpoint_path}/${epoch_id}/eval_model/`路径下，包含`__model__`和`__params__`两个文件。 -其中，`__model__`用于保存模型结构信息，`__params__`用于保存参数（parameters）信息。 -如果不需要保存评估模型，可以在定义Compressor对象时，将`save_eval_model`选项设置为False（默认为True）。 -运行命令为： -``` -python ../eval.py \ - --model_path ${checkpoint_path}/${epoch_id}/eval_model/ \ - --model_name __model__ \ - --params_name __params__ \ - -c ../../configs/yolov3_mobilenet_v1_voc.yml \ - -d "../../dataset/voc" -``` +### 保存断点（checkpoint） -## 预测 +蒸馏任务执行过程中会自动保存断点。如果需要从断点继续训练请用`-r`参数指定checkpoint路径，示例如下： -如果在配置文件中设置了`checkpoint_path`，并且在定义Compressor对象时指定了`prune_infer_model`选项，则每个epoch都会 -保存一个`inference model`。该模型是通过删除eval_program中多余的operators而得到的。 +```bash +python -u slim/distillation/distill.py \ +-c configs/yolov3_mobilenet_v1_voc.yml \ +-t configs/yolov3_r34_voc.yml \ +-r output/yolov3_mobilenet_v1_voc/10000 \ +--teacher_pretrained https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar +``` -该模型会保存在`${checkpoint_path}/${epoch_id}/eval_model/`路径下，包含`__model__.infer`和`__params__`两个文件。 -其中，`__model__.infer`用于保存模型结构信息，`__params__`用于保存参数（parameters）信息。 -更多关于`prune_infer_model`选项的介绍，请参考：[Compressor介绍](https://github.com/PaddlePaddle/models/blob/develop/PaddleSlim/docs/usage.md#121-%E5%A6%82%E4%BD%95%E6%94%B9%E5%86%99%E6%99%AE%E9%80%9A%E8%AE%AD%E7%BB%83%E8%84%9A%E6%9C%AC) -### python预测 -在脚本slim/infer.py中展示了如何使用fluid python API加载使用预测模型进行预测。 +## 评估 + +每隔`snap_shot_iter`步后会保存一个checkpoint模型可以用于评估，使用PaddleDetection目录下[tools/eval.py](../../tools/eval.py)评估脚本，并指定`weights`为训练得到的模型路径 运行命令为： -``` -python ../infer.py \ - --model_path ${checkpoint_path}/${epoch_id}/eval_model/ \ - --model_name __model__.infer \ - --params_name __params__ \ - -c ../../configs/yolov3_mobilenet_v1_voc.yml \ - --infer_dir ../../demo +```bash +export CUDA_VISIBLE_DEVICES=0 +python -u tools/eval.py -c configs/yolov3_mobilenet_v1_voc.yml \ + -o weights=output/yolov3_mobilenet_v1_voc/model_final \ ``` -### PaddleLite +## 预测 -该示例中产出的预测（inference）模型可以直接用PaddleLite进行加载使用。 -关于PaddleLite如何使用，请参考：[PaddleLite使用文档](https://github.com/PaddlePaddle/Paddle-Lite/wiki#%E4%BD%BF%E7%94%A8) +每隔`snap_shot_iter`步后保存的checkpoint模型也可以用于预测，使用PaddleDetection目录下[tools/infer.py](../../tools/infer.py)评估脚本，并指定`weights`为训练得到的模型路径 -## 示例结果 +### Python预测 ->当前release的结果并非超参调优后的最好结果，仅做示例参考，后续我们会优化当前结果。 +运行命令为： +``` +export CUDA_VISIBLE_DEVICES=0 +python -u tools/infer.py -c configs/yolov3_mobilenet_v1_voc.yml \ + --infer_img=demo/000000570688.jpg \ + --output_dir=infer_output/ \ + --draw_threshold=0.5 \ + -o weights=output/yolov3_mobilenet_v1_voc/model_final +``` -### MobileNetV1-YOLO-V3 +## 示例结果 -| FLOPS |Box AP| -|---|---| -|baseline|76.2 | -|蒸馏后|- | +### MobileNetV1-YOLO-V3-VOC +| FLOPS |输入尺寸|每张GPU图片个数|推理时间（fps）|Box AP|下载| +|:-:|:-:|:-:|:-:|:-:|:-:| +|baseline|608 |16|104.291|76.2|[下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar)| +|baseline|416 |16|-|76.7|[下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar)| +|baseline|320 |16|-|75.3|[下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar)| +|蒸馏后|608 |16|106.914|79.0|| +|蒸馏后|416 |16|-|78.2|| +|蒸馏后|320 |16|-|75.5|| -## FAQ +> 蒸馏后的结果用ResNet34-YOLO-V3做teacher，4GPU总batch_size64训练90000 iter得到 diff --git a/slim/distillation/compress.py b/slim/distillation/compress.py deleted file mode 100644 index 7c344aef6..000000000 --- a/slim/distillation/compress.py +++ /dev/null @@ -1,324 +0,0 @@ -# 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 multiprocessing -import numpy as np -from collections import deque, OrderedDict -from paddle.fluid.contrib.slim.core import Compressor -from paddle.fluid.framework import IrGraph - - -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 - -import sys -sys.path.append("../../") -from ppdet.core.workspace import load_config, merge_config, create -from ppdet.data.data_feed import create_reader -from ppdet.utils.eval_utils import parse_fetches, eval_results -from ppdet.utils.stats import TrainingStats -from ppdet.utils.cli import ArgsParser -from ppdet.utils.check import check_gpu -import ppdet.utils.checkpoint as checkpoint -from ppdet.modeling.model_input import create_feed - -import logging -FORMAT = '%(asctime)s-%(levelname)s: %(message)s' -logging.basicConfig(level=logging.INFO, format=FORMAT) -logger = logging.getLogger(__name__) - - -def eval_run(exe, compile_program, reader, keys, values, cls, test_feed): - """ - Run evaluation program, return program outputs. - """ - iter_id = 0 - results = [] - if len(cls) != 0: - values = [] - for i in range(len(cls)): - _, accum_map = cls[i].get_map_var() - cls[i].reset(exe) - values.append(accum_map) - - images_num = 0 - start_time = time.time() - has_bbox = 'bbox' in keys - for data in reader(): - data = test_feed.feed(data) - feed_data = {'image': data['image'], 'im_size': data['im_size']} - outs = exe.run(compile_program, - feed=feed_data, - fetch_list=[values[0]], - return_numpy=False) - outs.append(data['gt_box']) - outs.append(data['gt_label']) - outs.append(data['is_difficult']) - res = { - k: (np.array(v), v.recursive_sequence_lengths()) - for k, v in zip(keys, outs) - } - results.append(res) - if iter_id % 100 == 0: - logger.info('Test iter {}'.format(iter_id)) - iter_id += 1 - images_num += len(res['bbox'][1][0]) if has_bbox else 1 - logger.info('Test finish iter {}'.format(iter_id)) - - end_time = time.time() - fps = images_num / (end_time - start_time) - if has_bbox: - logger.info('Total number of images: {}, inference time: {} fps.'. - format(images_num, fps)) - else: - logger.info('Total iteration: {}, inference time: {} batch/s.'.format( - images_num, fps)) - - return results - - -def main(): - cfg = load_config(FLAGS.config) - if 'architecture' in cfg: - main_arch = cfg.architecture - else: - raise ValueError("'architecture' not specified in config file.") - - merge_config(FLAGS.opt) - if 'log_iter' not in cfg: - cfg.log_iter = 20 - - # check if set use_gpu=True in paddlepaddle cpu version - check_gpu(cfg.use_gpu) - - if cfg.use_gpu: - devices_num = fluid.core.get_cuda_device_count() - else: - devices_num = int( - os.environ.get('CPU_NUM', multiprocessing.cpu_count())) - - if 'train_feed' not in cfg: - train_feed = create(main_arch + 'TrainFeed') - else: - train_feed = create(cfg.train_feed) - - if 'eval_feed' not in cfg: - eval_feed = create(main_arch + 'EvalFeed') - else: - eval_feed = create(cfg.eval_feed) - - place = fluid.CUDAPlace(0) if cfg.use_gpu else fluid.CPUPlace() - exe = fluid.Executor(place) - - lr_builder = create('LearningRate') - optim_builder = create('OptimizerBuilder') - - # build program - model = create(main_arch) - _, train_feed_vars = create_feed(train_feed, True) - train_fetches = model.train(train_feed_vars) - loss = train_fetches['loss'] - lr = lr_builder() - opt = optim_builder(lr) - opt.minimize(loss) - #for v in fluid.default_main_program().list_vars(): - # if "py_reader" not in v.name and "double_buffer" not in v.name and "generated_var" not in v.name: - # print(v.name, v.shape) - - cfg.max_iters = 258 - train_reader = create_reader(train_feed, cfg.max_iters, FLAGS.dataset_dir) - - exe.run(fluid.default_startup_program()) - - # parse train fetches - train_keys, train_values, _ = parse_fetches(train_fetches) - train_keys.append('lr') - train_values.append(lr.name) - - train_fetch_list = [] - for k, v in zip(train_keys, train_values): - train_fetch_list.append((k, v)) - print("train_fetch_list: {}".format(train_fetch_list)) - - eval_prog = fluid.Program() - startup_prog = fluid.Program() - with fluid.program_guard(eval_prog, startup_prog): - with fluid.unique_name.guard(): - model = create(main_arch) - _, test_feed_vars = create_feed(eval_feed, True) - fetches = model.eval(test_feed_vars) - eval_prog = eval_prog.clone(True) - - eval_reader = create_reader(eval_feed, args_path=FLAGS.dataset_dir) - test_data_feed = fluid.DataFeeder(test_feed_vars.values(), 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'] - eval_keys, eval_values, eval_cls = parse_fetches(fetches, eval_prog, - extra_keys) - - eval_fetch_list = [] - for k, v in zip(eval_keys, eval_values): - eval_fetch_list.append((k, v)) - print("eval_fetch_list: {}".format(eval_fetch_list)) - - exe.run(startup_prog) - checkpoint.load_params(exe, - fluid.default_main_program(), cfg.pretrain_weights) - - best_box_ap_list = [] - - def eval_func(program, scope): - results = eval_run(exe, program, eval_reader, eval_keys, eval_values, - eval_cls, test_data_feed) - - resolution = None - is_bbox_normalized = False - if 'mask' in results[0]: - resolution = model.mask_head.resolution - box_ap_stats = eval_results(results, eval_feed, cfg.metric, - cfg.num_classes, resolution, - is_bbox_normalized, FLAGS.output_eval) - if len(best_box_ap_list) == 0: - best_box_ap_list.append(box_ap_stats[0]) - elif box_ap_stats[0] > best_box_ap_list[0]: - best_box_ap_list[0] = box_ap_stats[0] - logger.info("Best test box ap: {}".format(best_box_ap_list[0])) - return best_box_ap_list[0] - - test_feed = [('image', test_feed_vars['image'].name), - ('im_size', test_feed_vars['im_size'].name)] - - teacher_cfg = load_config(FLAGS.teacher_config) - teacher_arch = teacher_cfg.architecture - teacher_programs = [] - teacher_program = fluid.Program() - teacher_startup_program = fluid.Program() - with fluid.program_guard(teacher_program, teacher_startup_program): - with fluid.unique_name.guard('teacher_'): - teacher_feed_vars = OrderedDict() - for name, var in train_feed_vars.items(): - teacher_feed_vars[name] = teacher_program.global_block( - )._clone_variable( - var, force_persistable=False) - model = create(teacher_arch) - train_fetches = model.train(teacher_feed_vars) - #print("="*50+"teacher_model_params"+"="*50) - #for v in teacher_program.list_vars(): - # print(v.name, v.shape) - #return - - exe.run(teacher_startup_program) - assert FLAGS.teacher_pretrained and os.path.exists( - FLAGS.teacher_pretrained - ), "teacher_pretrained should be set when teacher_model is not None." - - def if_exist(var): - return os.path.exists(os.path.join(FLAGS.teacher_pretrained, var.name)) - - fluid.io.load_vars( - exe, - FLAGS.teacher_pretrained, - main_program=teacher_program, - predicate=if_exist) - - teacher_programs.append(teacher_program.clone(for_test=True)) - - com = Compressor( - place, - fluid.global_scope(), - fluid.default_main_program(), - train_reader=train_reader, - train_feed_list=[(key, value.name) - for key, value in train_feed_vars.items()], - train_fetch_list=train_fetch_list, - eval_program=eval_prog, - eval_reader=eval_reader, - eval_feed_list=test_feed, - eval_func={'map': eval_func}, - eval_fetch_list=eval_fetch_list[0:1], - save_eval_model=True, - prune_infer_model=[["image", "im_size"], ["multiclass_nms_0.tmp_0"]], - teacher_programs=teacher_programs, - train_optimizer=None, - distiller_optimizer=opt, - log_period=20) - com.config(FLAGS.slim_file) - com.run() - - -if __name__ == '__main__': - parser = ArgsParser() - parser.add_argument( - "-t", - "--teacher_config", - default=None, - type=str, - help="Config file of teacher architecture.") - parser.add_argument( - "-s", - "--slim_file", - default=None, - type=str, - help="Config file of PaddleSlim.") - parser.add_argument( - "-r", - "--resume_checkpoint", - default=None, - type=str, - help="Checkpoint path for resuming training.") - parser.add_argument( - "--eval", - action='store_true', - default=False, - help="Whether to perform evaluation in train") - parser.add_argument( - "--teacher_pretrained", - default=None, - type=str, - help="Whether to use pretrained model.") - 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") - FLAGS = parser.parse_args() - main() diff --git a/slim/distillation/distill.py b/slim/distillation/distill.py new file mode 100644 index 000000000..f313dd4cf --- /dev/null +++ b/slim/distillation/distill.py @@ -0,0 +1,289 @@ +# 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 numpy as np +from collections import OrderedDict +from paddleslim.dist.single_distiller import merge, l2_loss + +from paddle import fluid +from ppdet.core.workspace import load_config, merge_config, create +from ppdet.data.reader import create_reader +from ppdet.utils.eval_utils import parse_fetches, eval_results, eval_run +from ppdet.utils.stats import TrainingStats +from ppdet.utils.cli import ArgsParser +from ppdet.utils.check import check_gpu +import ppdet.utils.checkpoint as checkpoint + +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 + cfg = load_config(FLAGS.config) + if 'architecture' in cfg: + main_arch = cfg.architecture + else: + raise ValueError("'architecture' not specified in config file.") + + merge_config(FLAGS.opt) + if 'log_iter' not in cfg: + cfg.log_iter = 20 + + # check if set use_gpu=True in paddlepaddle cpu version + check_gpu(cfg.use_gpu) + + if cfg.use_gpu: + devices_num = fluid.core.get_cuda_device_count() + else: + devices_num = int(os.environ.get('CPU_NUM', 1)) + + if 'FLAGS_selected_gpus' in env: + device_id = int(env['FLAGS_selected_gpus']) + else: + device_id = 0 + place = fluid.CUDAPlace(device_id) if cfg.use_gpu else fluid.CPUPlace() + exe = fluid.Executor(place) + + # build program + model = create(main_arch) + inputs_def = cfg['TrainReader']['inputs_def'] + train_feed_vars, train_loader = model.build_inputs(**inputs_def) + train_fetches = model.train(train_feed_vars) + loss = train_fetches['loss'] + # get all student variables + student_vars = [] + for v in fluid.default_main_program().list_vars(): + try: + student_vars.append((v.name, v.shape)) + except: + pass + # uncomment the following lines to print all student variables + # print("="*50 + "student_model_vars" + "="*50) + # print(student_vars) + + eval_prog = fluid.Program() + with fluid.program_guard(eval_prog, fluid.default_startup_program()): + with fluid.unique_name.guard(): + model = create(main_arch) + inputs_def = cfg['EvalReader']['inputs_def'] + test_feed_vars, eval_loader = model.build_inputs(**inputs_def) + fetches = model.eval(test_feed_vars) + eval_prog = eval_prog.clone(True) + + 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_bbox', 'gt_class', 'is_difficult'] + eval_keys, eval_values, eval_cls = parse_fetches(fetches, eval_prog, + extra_keys) + + teacher_cfg = load_config(FLAGS.teacher_config) + teacher_arch = teacher_cfg.architecture + teacher_program = fluid.Program() + teacher_startup_program = fluid.Program() + + with fluid.program_guard(teacher_program, teacher_startup_program): + with fluid.unique_name.guard(): + teacher_feed_vars = OrderedDict() + for name, var in train_feed_vars.items(): + teacher_feed_vars[name] = teacher_program.global_block( + )._clone_variable( + var, force_persistable=False) + model = create(teacher_arch) + train_fetches = model.train(teacher_feed_vars) + teacher_loss = train_fetches['loss'] + + # get all teacher variables + teacher_vars = [] + for v in teacher_program.list_vars(): + try: + teacher_vars.append((v.name, v.shape)) + except: + pass + # uncomment the following lines to print all teacher variables + # print("="*50 + "teacher_model_vars" + "="*50) + # print(teacher_vars) + + exe.run(teacher_startup_program) + assert FLAGS.teacher_pretrained, "teacher_pretrained should be set" + checkpoint.load_params(exe, teacher_program, FLAGS.teacher_pretrained) + teacher_program = teacher_program.clone(for_test=True) + + cfg = load_config(FLAGS.config) + data_name_map = { + 'image': 'image', + 'gt_bbox': 'gt_bbox', + 'gt_class': 'gt_class', + 'gt_score': 'gt_score' + } + distill_prog = merge(teacher_program, + fluid.default_main_program(), data_name_map, place) + + distill_weight = 100 + distill_pairs = [['teacher_conv2d_6.tmp_1', 'conv2d_20.tmp_1'], + ['teacher_conv2d_14.tmp_1', 'conv2d_28.tmp_1'], + ['teacher_conv2d_22.tmp_1', 'conv2d_36.tmp_1']] + + def l2_distill(pairs, weight): + """ + Add l2 distillation losses composed of multi pairs of feature maps, + each pair of feature maps is the input of teacher and student's + yolov3_loss respectively + """ + loss = [] + for pair in pairs: + loss.append(l2_loss(pair[0], pair[1])) + loss = fluid.layers.sum(loss) + weighted_loss = loss * weight + return weighted_loss + + distill_loss = l2_distill(distill_pairs, distill_weight) + loss = distill_loss + loss + lr_builder = create('LearningRate') + optim_builder = create('OptimizerBuilder') + lr = lr_builder() + opt = optim_builder(lr) + opt.minimize(loss) + + exe.run(fluid.default_startup_program()) + checkpoint.load_params(exe, + fluid.default_main_program(), cfg.pretrain_weights) + + build_strategy = fluid.BuildStrategy() + build_strategy.fuse_all_reduce_ops = False + build_strategy.fuse_all_optimizer_ops = False + build_strategy.fuse_elewise_add_act_ops = True + # only enable sync_bn in multi GPU devices + sync_bn = getattr(model.backbone, 'norm_type', None) == 'sync_bn' + build_strategy.sync_batch_norm = sync_bn and devices_num > 1 \ + and cfg.use_gpu + + exec_strategy = fluid.ExecutionStrategy() + # iteration number when CompiledProgram tries to drop local execution scopes. + # Set it to be 1 to save memory usages, so that unused variables in + # local execution scopes can be deleted after each iteration. + exec_strategy.num_iteration_per_drop_scope = 1 + + parallel_main = fluid.CompiledProgram(distill_prog).with_data_parallel( + loss_name=loss.name, + build_strategy=build_strategy, + exec_strategy=exec_strategy) + + compiled_eval_prog = fluid.compiler.CompiledProgram(eval_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 FLAGS.resume_checkpoint: + checkpoint.load_checkpoint(exe, distill_prog, FLAGS.resume_checkpoint) + start_iter = checkpoint.global_step() + elif cfg.pretrain_weights and fuse_bn and not ignore_params: + checkpoint.load_and_fusebn(exe, distill_prog, cfg.pretrain_weights) + elif cfg.pretrain_weights: + checkpoint.load_params( + exe, + distill_prog, + cfg.pretrain_weights, + ignore_params=ignore_params) + + train_reader = create_reader(cfg.TrainReader, (cfg.max_iters - start_iter) * + devices_num, cfg) + train_loader.set_sample_list_generator(train_reader, place) + # 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() + map_type = cfg.map_type if 'map_type' in cfg else '11point' + best_box_ap_list = [0.0, 0] #[map, iter] + cfg_name = os.path.basename(FLAGS.config).split('.')[0] + save_dir = os.path.join(cfg.save_dir, cfg_name) + + train_loader.start() + for step_id in range(start_iter, cfg.max_iters): + teacher_loss_np, distill_loss_np, loss_np, lr_np = exe.run( + parallel_main, + fetch_list=[ + 'teacher_' + teacher_loss.name, distill_loss.name, loss.name, + lr.name + ]) + if step_id % cfg.log_iter == 0: + logger.info( + "step {} lr {:.6f}, loss {:.6f}, distill_loss {:.6f}, teacher_loss {:.6f}". + format(step_id, lr_np[0], loss_np[0], distill_loss_np[0], + teacher_loss_np[0])) + if step_id % cfg.snapshot_iter == 0 and step_id != 0 or step_id == cfg.max_iters - 1: + save_name = str( + step_id) if step_id != cfg.max_iters - 1 else "model_final" + checkpoint.save(exe, distill_prog, + os.path.join(save_dir, save_name)) + # eval + results = eval_run(exe, compiled_eval_prog, eval_loader, eval_keys, + eval_values, eval_cls) + resolution = None + box_ap_stats = eval_results(results, cfg.metric, cfg.num_classes, + resolution, is_bbox_normalized, + FLAGS.output_eval, map_type, + cfg['EvalReader']['dataset']) + + if box_ap_stats[0] > best_box_ap_list[0]: + best_box_ap_list[0] = box_ap_stats[0] + best_box_ap_list[1] = step_id + checkpoint.save(exe, distill_prog, + os.path.join("./", "best_model")) + logger.info("Best test box ap: {}, in step: {}".format( + best_box_ap_list[0], best_box_ap_list[1])) + train_loader.reset() + + +if __name__ == '__main__': + parser = ArgsParser() + parser.add_argument( + "-r", + "--resume_checkpoint", + default=None, + type=str, + help="Checkpoint path for resuming training.") + parser.add_argument( + "-t", + "--teacher_config", + default=None, + type=str, + help="Config file of teacher architecture.") + parser.add_argument( + "--teacher_pretrained", + default=None, + type=str, + help="Whether to use pretrained model.") + parser.add_argument( + "--output_eval", + default=None, + type=str, + help="Evaluation directory, default is current directory.") + FLAGS = parser.parse_args() + main() diff --git a/slim/distillation/run.sh b/slim/distillation/run.sh deleted file mode 100644 index a5497bdce..000000000 --- a/slim/distillation/run.sh +++ /dev/null @@ -1,47 +0,0 @@ -#!/usr/bin/env bash - -# download pretrain model -root_url="https://paddlemodels.bj.bcebos.com/object_detection" -yolov3_r34_voc="yolov3_r34_voc.tar" -pretrain_dir='./pretrain' - -if [ ! -d ${pretrain_dir} ]; then - mkdir ${pretrain_dir} -fi - -cd ${pretrain_dir} - -if [ ! -f ${yolov3_r34_voc} ]; then - wget ${root_url}/${yolov3_r34_voc} - tar xf ${yolov3_r34_voc} -fi -cd - - -# enable GC strategy -export FLAGS_fast_eager_deletion_mode=1 -export FLAGS_eager_delete_tensor_gb=0.0 - -# for distillation -#----------------- -export CUDA_VISIBLE_DEVICES=0,1,2,3 - - -# Fixing name conflicts in distillation -cd ${pretrain_dir}/yolov3_r34_voc -for files in $(ls teacher_*) - do mv $files ${files#*_} -done -for files in $(ls *) - do mv $files "teacher_"$files -done -cd - - -python -u compress.py \ --c ../../configs/yolov3_mobilenet_v1_voc.yml \ --t yolov3_resnet34.yml \ --s yolov3_mobilenet_v1_yolov3_resnet34_distillation.yml \ --o YoloTrainFeed.batch_size=64 \ --d ../../dataset/voc \ ---teacher_pretrained ./pretrain/yolov3_r34_voc \ -> yolov3_distallation.log 2>&1 & -tailf yolov3_distallation.log diff --git a/slim/distillation/yolov3_mobilenet_v1_yolov3_resnet34_distillation.yml b/slim/distillation/yolov3_mobilenet_v1_yolov3_resnet34_distillation.yml deleted file mode 100644 index 6a2a5a257..000000000 --- a/slim/distillation/yolov3_mobilenet_v1_yolov3_resnet34_distillation.yml +++ /dev/null @@ -1,18 +0,0 @@ -version: 1.0 -distillers: - l2_distiller: - class: 'L2Distiller' - teacher_feature_map: 'teacher_teacher_conv2d_1.tmp_0' - student_feature_map: 'conv2d_15.tmp_0' - distillation_loss_weight: 1 -strategies: - distillation_strategy: - class: 'DistillationStrategy' - distillers: ['l2_distiller'] - start_epoch: 0 - end_epoch: 270 -compressor: - epoch: 271 - checkpoint_path: './checkpoints/' - strategies: - - distillation_strategy diff --git a/slim/distillation/yolov3_resnet34.yml b/slim/distillation/yolov3_resnet34.yml deleted file mode 100644 index c04bdde9f..000000000 --- a/slim/distillation/yolov3_resnet34.yml +++ /dev/null @@ -1,34 +0,0 @@ -architecture: YOLOv3 -log_smooth_window: 20 -metric: VOC -map_type: 11point -num_classes: 20 -weight_prefix_name: teacher_ - -YOLOv3: - backbone: ResNet - yolo_head: YOLOv3Head - -ResNet: - norm_type: sync_bn - freeze_at: 0 - freeze_norm: false - norm_decay: 0. - depth: 34 - feature_maps: [3, 4, 5] - -YOLOv3Head: - anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]] - anchors: [[10, 13], [16, 30], [33, 23], - [30, 61], [62, 45], [59, 119], - [116, 90], [156, 198], [373, 326]] - norm_decay: 0. - ignore_thresh: 0.7 - label_smooth: false - nms: - background_label: -1 - keep_top_k: 100 - nms_threshold: 0.45 - nms_top_k: 1000 - normalized: false - score_threshold: 0.01 -- GitLab