官方的yolov5n模型我自己用opt转换后使用报错

[leslie2046]

为使您的问题得到快速解决，在建立 Issue 前，请您先通过如下方式搜索是否有相似问题: 历史 issue, FAQ 文档, 官方文档

建立 issue 时，为快速解决问题，请您根据使用情况给出如下信息：

• 标题：简洁、精准描述您的问题，例如“ssd 模型转换报错”
• 版本、环境信息：
     1）Paddle Lite 版本：请提供您的 Paddle Lite 版本号（如v2.10）或 CommitID
  paddle lite:v2.11
  paddlepaddle-gpu 2.4.2
     2）Host 环境：请描述 Host 系统类型、OS 版本，如 Mac OS 10.14、Ubuntu 18.04
  host centos
• 模型信息
     1）模型名称 yolov5n
  2）模型链接
  https://paddledet.bj.bcebos.com/models/yolov5_n_300e_coco.pdparams
• 复现信息：提供 OPT 使用命令或方法，用于复现问题

1. python tools/export_model.py -c configs/yolov5/yolov5_n_300e_coco.yml
   -o weights=output/yolov5_n_300e_coco/yolov5_n_300e_coco.pdparams --output_dir=output_inference
2. ./opt_linux --valid_targets=arm,opencl
   --model_file=output_inference/yolov5_n_300e_coco/model.pdmodel
   --param_file=output_inference/yolov5_n_300e_coco/model.pdiparams
   --optimize_out=output_inference/yolov5_n_300e_coco/model

• 问题描述：请详细描述您的问题，同步贴出报错信息、日志/代码关键片段
  我是在这个demo里使用的，https://github.com/PaddlePaddle/Paddle-Lite-Demo/tree/develop/object_detection/android/app/cxx/yolov5n_detection_demo
  我用的opencl版本的paddle Lite库https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.11/inference_lite_lib.armv8.clang.with_exception.with_extra.with_cv.opencl.tar.gz
  用的demo里自带的model.nb可以正常运行。
  但是用上述方法自己转换和opt优化后的model.nb却无法在demo中运行起来，会报错 terminating with uncaught exception of type std::exception: std::exception

[leslie2046]

python tools/export_model.py -c configs/yolov5/yolov5_n_300e_coco.yml \

-o weights=output/yolov5_n_300e_coco/yolov5_n_300e_coco.pdparams --output_dir=output_inference
/home/czr/anaconda3/envs/paddleyolo2/lib/python3.9/site-packages/setuptools/command/easy_install.py:41: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html
import pkg_resources
/home/czr/anaconda3/envs/paddleyolo2/lib/python3.9/site-packages/pkg_resources/init.py:3154: DeprecationWarning: Deprecated call to pkg_resources.declare_namespace('google').
Implementing implicit namespace packages (as specified in PEP 420) is preferred to pkg_resources.declare_namespace. See https://setuptools.pypa.io/en/latest/references/keywords.html#keyword-namespace-packages
declare_namespace(pkg)
[10/30 14:40:55] ppdet.engine INFO: Model Params : [1.872157] M.
[10/30 14:40:55] ppdet.utils.checkpoint INFO: Finish loading model weights: output/yolov5_n_300e_coco/yolov5_n_300e_coco.pdparams
loading annotations into memory...
Done (t=0.65s)
creating index...
index created!
[10/30 14:40:56] ppdet.engine INFO: Export inference config file to output_inference/yolov5_n_300e_coco/infer_cfg.yml
[10/30 14:41:01] ppdet.engine INFO: Export model and saved in output_inference/yolov5_n_300e_coco

[leslie2046]

./opt_linux --valid_targets=arm,opencl --model_file=output_inference/yolov5_n_300e_coco/model.pdmodel --param_file=output_inference/yolov5_n_300e_coco/model.pdiparams --optimize_out=output_inference/yolov5_n_300e_coco/model
Loading topology data from output_inference/yolov5_n_300e_coco/model.pdmodel
Loading params data from output_inference/yolov5_n_300e_coco/model.pdiparams

1. Model is successfully loaded!
2. Model is optimized and saved into output_inference/yolov5_n_300e_coco/model.nb successfully

[leslie2046]

yolov5_n_300e_coco.zip
这是model.nb

[hong19860320]

可以尝试使用 release note 最新的 2.14-rc opt 工具和 Paddle Lite 库试试哈~
https://github.com/PaddlePaddle/Paddle-Lite/releases

[leslie2046]

可以尝试使用 release note 最新的 2.14-rc opt 工具和 Paddle Lite 库试试哈~ https://github.com/PaddlePaddle/Paddle-Lite/releases

python tools/export_model.py -c configs/yolov5/yolov5_n_300e_coco.yml -o weights=output/yolov5_n_300e_coco/yolov5_n_300e_coco.pdparams --output_dir=output_inference
./opt_linux_x86_2.14rc --valid_targets=opencl,arm --model_file=output_inference/yolov5_n_300e_coco/model.pdmodel --param_file=output_inference/yolov5_n_300e_coco/model.pdiparams --optimize_out=output_inference/yolov5_n_300e_coco/model --optimize_out_type=naive_buffer
然后PaddleLiteDemo里也换成了v2.14-rc版本的库，依然报错

[leslie2046]

现在能跑起来了，但是会时不时的卡几秒钟，卡顿的时候预测耗时达到好几秒

[leslie2046]

#10585