Backbone:PPLCNetV2_base_ShiTu 识别模式转换失败

[hepei119]

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：paddlelite=2.13rc0 2) PaddleClas 版本：2.5.1_release 3) OS 版本 22.04.1-Ubuntu

• 模型信息 1) 模型链接: inference_model.zip 2) error.log

• 复现信息： 指令：paddle_lite_opt --valid_targets=arm --model_file=/home/xnzn/workspace3/Paddle-Lite/inference6/inference.pdmodel --param_file=/home/xnzn/workspace3/Paddle-Lite/inference6/inference.pdiparams --optimize_out=/home/xnzn/workspace3/Paddle-Lite/inference6/rec

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 [6 4/12 7:47:54.459 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/12 7:47:54.459 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/12 7:47:54.459 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/12 7:47:54.459 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/12 7:47:54.459 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/12 7:47:54.459 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/12 7:47:54.459 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/12 7:47:54.459 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation Segmentation fault (core dumped)

[lishicheng1996]

已收到您的issue，等内部同学看一下

[lishicheng1996]

您的log里没有看到报错诶，方便贴一下带报错的部分吗

[hepei119]

设置export GLOG_v=9 只有一段 Segmentation fault (core dumped)

[lishicheng1996]

试试把右边的x直接reshape到[1, 256, 14, 14]看看

[hepei119]

我将is_repped改成False得到的网络结构：

模型文件连接： inference12.zip

但是在opt执行依旧提示 [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] depthwise_conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] depthwise_conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] depthwise_conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] depthwise_conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] depthwise_conv2d [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation [6 4/14 6:45: 1.194 ...ite/lite/core/optimizer/mir/ssa_graph.cc:30 CheckBidirectionalConnection] fusion_elementwise_add_activation Segmentation fault (core dumped) error.log

[hong19860320]

通过基于模型裁剪工具 https://github.com/hong19860320/PaddleInference-generic-demo/blob/develop/tools/model_utilities/split_model.py 和二分法裁剪模型方法，定位到这个算子有点问题，可能 Paddle 主框架调整了该算子定义，Paddle Lite 暂时没有适配导致的。

[ddchenhao66]

您可以试试拉develop最新分支编译运行，这边试过是能跑通的～

[hepei119]

感谢用develop验证了没有问题