From 0a11e56ba9bbe2211815e22b03d0447dae28a8e4 Mon Sep 17 00:00:00 2001 From: wangguanzhong Date: Thu, 24 Oct 2019 10:18:19 +0800 Subject: [PATCH] add export_model doc (#3709) * add export_model doc --- README_cn.md | 1 + docs/EXPORT_MODEL.md | 38 ++++++++++++ docs/GETTING_STARTED.md | 14 ++--- docs/GETTING_STARTED_cn.md | 13 ---- tools/export_model.py | 119 +++++++++++++++++++++++++++++++++++++ tools/infer.py | 50 ---------------- 6 files changed, 164 insertions(+), 71 deletions(-) create mode 100644 docs/EXPORT_MODEL.md create mode 100644 tools/export_model.py diff --git a/README_cn.md b/README_cn.md index 293b65959..7793d6495 100644 --- a/README_cn.md +++ b/README_cn.md @@ -75,6 +75,7 @@ PaddleDetection的目的是为工业界和学术界提供丰富、易用的目 ## 推理部署 +- [模型导出教程](docs/EXPORT_MODEL.md) - [C++推理部署](inference/README.md) ## Benchmark diff --git a/docs/EXPORT_MODEL.md b/docs/EXPORT_MODEL.md new file mode 100644 index 000000000..7c6dff1f0 --- /dev/null +++ b/docs/EXPORT_MODEL.md @@ -0,0 +1,38 @@ +# 模型导出 + +训练得到一个满足要求的模型后，如果想要将该模型接入到C++预测库或者Serving服务，需要通过`tools/export_model.py`导出该模型。 + +## 启动参数说明 + +| FLAG | 用途 | 默认值 | 备注 | +|:--------------:|:--------------:|:------------:|:-----------------------------------------:| +| -c | 指定配置文件 | None | | +| --output_dir | 模型保存路径 | `./output` | 模型默认保存在`output/配置文件名/`路径下 | + +## 使用示例 + +使用[训练/评估/推断](GETTING_STARTED_cn.md)中训练得到的模型进行试用，脚本如下 + +```bash +# export model for RCNN +python tools/export_model.py -c configs/faster_rcnn_r50_1x.yml \ + --output_dir=./inference_model \ + -o weights=output/faster_rcnn_r50_1x/model_final \ + FasterRCNNTestFeed.image_shape=[3,800,1333] + +# export model for YOLOv3 +python tools/export_model.py -c configs/yolov3_darknet.yml \ + --output_dir=./inference_model \ + -o weights=output/yolov3_darknet/model_final \ + YoloTestFeed.image_shape=[3,800,1333] + +# export model for SSD +python tools/export_model.py -c configs/ssd/ssd_mobilenet_v1_voc.yml \ + --output_dir=./inference_model \ + -o weights=output/ssd_mobilenet_v1_voc/model_final \ + SSDTestFeed.image_shape=[3,300,300] + +``` + +- 预测模型会导出到`output/faster_rcnn_r50_1x`目录下，模型名和参数名分别为`__model__`和`__params__`。 +- 通过`image_shape`修改保存模型中的图片大小。使用Fluid-TensorRT进行预测时，由于TensorRT仅支持定长输入，需要将输入图片大小与`image_shape`保持一致。 diff --git a/docs/GETTING_STARTED.md b/docs/GETTING_STARTED.md index 82ab2e98e..199b343ed 100644 --- a/docs/GETTING_STARTED.md +++ b/docs/GETTING_STARTED.md @@ -40,7 +40,6 @@ list below can be viewed by `--help` | --json_eval | eval | Whether to evaluate with already existed bbox.json or mask.json | False | json path is set in `--output_eval` | | --output_dir | infer | Directory for storing the output visualization files | `./output` | `--output_dir output` | | --draw_threshold | infer | Threshold to reserve the result for visualization | 0.5 | `--draw_threshold 0.7` | -| --save\_inference_model | infer | Whether to save inference model in output_dir | False | save_inference_model is saved in `--output_dir` | | --infer_dir | infer | Directory for images to perform inference on | None | | | --infer_img | infer | Image path | None | higher priority over --infer_dir | | --use_tb | train/infer | Whether to record the data with [tb-paddle](https://github.com/linshuliang/tb-paddle), so as to display in Tensorboard | False | | @@ -149,17 +148,16 @@ moment, but it is a planned feature Different thresholds will produce different results depending on the calculation of [NMS](https://ieeexplore.ieee.org/document/1699659). -- Save inference model +- Export model ```bash - export CUDA_VISIBLE_DEVICES=0 - python tools/infer.py -c configs/faster_rcnn_r50_1x.yml \ - --infer_img=demo/000000570688.jpg \ - --save_inference_model + python tools/export_model.py -c configs/faster_rcnn_r50_1x.yml \ + --output_dir=inference_model \ + -o weights=output/faster_rcnn_r50_1x/model_final \ + FasterRCNNTestFeed.image_shape=[3,800,1333] ``` - Save inference model by set `--save_inference_model`, which can be loaded by PaddlePaddle predict library. - + Save inference model `tools/export_model.py`, which can be loaded by PaddlePaddle predict library. ## FAQ diff --git a/docs/GETTING_STARTED_cn.md b/docs/GETTING_STARTED_cn.md index a32bfa7a8..b5dd60410 100644 --- a/docs/GETTING_STARTED_cn.md +++ b/docs/GETTING_STARTED_cn.md @@ -37,7 +37,6 @@ python tools/infer.py -c configs/faster_rcnn_r50_1x.yml --infer_img=demo/0000005 | --json_eval | eval | 是否通过已存在的bbox.json或者mask.json进行评估 | False | json文件路径在`--output_eval`中设置 | | --output_dir | infer | 输出推断后可视化文件 | `./output` | `--output_dir output` | | --draw_threshold | infer | 可视化时分数阈值 | 0.5 | `--draw_threshold 0.7` | -| --save\_inference_model | infer | 是否保存预测模型 | False | 预测模型保存路径为`--output_dir`设置的路径 | | --infer_dir | infer | 用于推断的图片文件夹路径 | None | | | --infer_img | infer | 用于推断的图片路径 | None | 相较于`--infer_dir`具有更高优先级 | | --use_tb | train/infer | 是否使用[tb-paddle](https://github.com/linshuliang/tb-paddle)记录数据，进而在TensorBoard中显示 | False | | @@ -145,18 +144,6 @@ python -m paddle.distributed.launch --selected_gpus 0,1,2,3,4,5,6,7 tools/train. `--draw_threshold` 是个可选参数. 根据 [NMS](https://ieeexplore.ieee.org/document/1699659) 的计算， 不同阈值会产生不同的结果。如果用户需要对自定义路径的模型进行推断，可以设置`-o weights`指定模型路径。 -- 保存推断模型 - - ```bash - # GPU推断 - export CUDA_VISIBLE_DEVICES=0 - python -u tools/infer.py -c configs/faster_rcnn_r50_1x.yml --infer_img=demo/000000570688.jpg \ - --save_inference_model - ``` - - 通过设置`--save_inference_model`保存可供PaddlePaddle预测库加载的推断模型。 - - ## FAQ **Q:** 为什么我使用单GPU训练loss会出`NaN`?
diff --git a/tools/export_model.py b/tools/export_model.py new file mode 100644 index 000000000..f83371f20 --- /dev/null +++ b/tools/export_model.py @@ -0,0 +1,119 @@ +# 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 + +from paddle import fluid + +from ppdet.core.workspace import load_config, merge_config, create +from ppdet.modeling.model_input import create_feed +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 prune_feed_vars(feeded_var_names, target_vars, prog): + """ + Filter out feed variables which are not in program, + pruned feed variables are only used in post processing + on model output, which are not used in program, such + as im_id to identify image order, im_shape to clip bbox + in image. + """ + exist_var_names = [] + prog = prog.clone() + prog = prog._prune(targets=target_vars) + global_block = prog.global_block() + for name in feeded_var_names: + try: + v = global_block.var(name) + exist_var_names.append(str(v.name)) + except Exception: + logger.info('save_inference_model pruned unused feed ' + 'variables {}'.format(name)) + pass + return exist_var_names + + +def save_infer_model(FLAGS, exe, feed_vars, test_fetches, infer_prog): + cfg_name = os.path.basename(FLAGS.config).split('.')[0] + save_dir = os.path.join(FLAGS.output_dir, cfg_name) + feed_var_names = [var.name for var in feed_vars.values()] + target_vars = list(test_fetches.values()) + feed_var_names = prune_feed_vars(feed_var_names, target_vars, infer_prog) + logger.info("Export inference model to {}, input: {}, output: " + "{}...".format(save_dir, feed_var_names, + [str(var.name) for var in target_vars])) + fluid.io.save_inference_model( + save_dir, + feeded_var_names=feed_var_names, + target_vars=target_vars, + executor=exe, + main_program=infer_prog, + params_filename="__params__") + + +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 'test_feed' not in cfg: + test_feed = create(main_arch + 'TestFeed') + else: + test_feed = create(cfg.test_feed) + + # Use CPU for exporting inference model instead of GPU + place = fluid.CPUPlace() + exe = fluid.Executor(place) + + model = create(main_arch) + + startup_prog = fluid.Program() + infer_prog = fluid.Program() + with fluid.program_guard(infer_prog, startup_prog): + with fluid.unique_name.guard(): + _, feed_vars = create_feed(test_feed, use_pyreader=False) + test_fetches = model.test(feed_vars) + infer_prog = infer_prog.clone(True) + + exe.run(startup_prog) + checkpoint.load_params(exe, infer_prog, cfg.weights) + + save_infer_model(FLAGS, exe, feed_vars, test_fetches, infer_prog) + + +if __name__ == '__main__': + parser = ArgsParser() + parser.add_argument( + "--output_dir", + type=str, + default="output", + help="Directory for storing the output model files.") + FLAGS = parser.parse_args() + main() diff --git a/tools/infer.py b/tools/infer.py index cec4aaa91..ed1081484 100644 --- a/tools/infer.py +++ b/tools/infer.py @@ -96,48 +96,6 @@ def get_test_images(infer_dir, infer_img): return images -def prune_feed_vars(feeded_var_names, target_vars, prog): - """ - Filter out feed variables which are not in program, - pruned feed variables are only used in post processing - on model output, which are not used in program, such - as im_id to identify image order, im_shape to clip bbox - in image. - """ - exist_var_names = [] - prog = prog.clone() - prog = prog._prune(targets=target_vars) - global_block = prog.global_block() - for name in feeded_var_names: - try: - v = global_block.var(name) - exist_var_names.append(str(v.name)) - except Exception: - logger.info('save_inference_model pruned unused feed ' - 'variables {}'.format(name)) - pass - return exist_var_names - - -def save_infer_model(FLAGS, exe, feed_vars, test_fetches, infer_prog): - cfg_name = os.path.basename(FLAGS.config).split('.')[0] - save_dir = os.path.join(FLAGS.output_dir, cfg_name) - feeded_var_names = [var.name for var in feed_vars.values()] - target_vars = list(test_fetches.values()) - feeded_var_names = prune_feed_vars(feeded_var_names, target_vars, - infer_prog) - logger.info("Save inference model to {}, input: {}, output: " - "{}...".format(save_dir, feeded_var_names, - [str(var.name) for var in target_vars])) - fluid.io.save_inference_model( - save_dir, - feeded_var_names=feeded_var_names, - target_vars=target_vars, - executor=exe, - main_program=infer_prog, - params_filename="__params__") - - def main(): cfg = load_config(FLAGS.config) @@ -180,9 +138,6 @@ def main(): if cfg.weights: checkpoint.load_params(exe, infer_prog, cfg.weights) - if FLAGS.save_inference_model: - save_infer_model(FLAGS, exe, feed_vars, test_fetches, infer_prog) - # parse infer fetches assert cfg.metric in ['COCO', 'VOC', 'WIDERFACE'], \ "unknown metric type {}".format(cfg.metric) @@ -300,11 +255,6 @@ if __name__ == '__main__': type=float, default=0.5, help="Threshold to reserve the result for visualization.") - parser.add_argument( - "--save_inference_model", - action='store_true', - default=False, - help="Save inference model in output_dir if True.") parser.add_argument( "--use_tb", type=bool, -- GitLab