这是啥情况,checker时只报个这错误Segmentation fault,啥也不说,从哪错的也不知道

2021-11-29 15:23:11,101 INFO Start hb_mapper…

2021-11-29 15:23:11,101 INFO log will be stored in /data/horizon_xj3_open_explorer_v1.5.6_20210930/ddk/samples/ai_toolchain/horizon_model_convert_sample/mytest/./retinaface_onnx_checker.log

2021-11-29 15:23:11,102 INFO hbdk version 3.24.4

2021-11-29 15:23:11,102 INFO horizon_nn version 0.11.2

2021-11-29 15:23:11,103 INFO hb_mapper version 1.3.53

2021-11-29 15:23:11,116 INFO Model type: onnx

2021-11-29 15:23:11,117 INFO input names

2021-11-29 15:23:11,117 INFO input shapes {}

2021-11-29 15:23:11,118 INFO Begin model checking…

2021-11-29 15:23:11,122 INFO [Mon Nov 29 15:23:11 2021] Start to Horizon NN Model Convert.

2021-11-29 15:23:11,123 INFO The input parameter is not specified, convert with default parameters.

2021-11-29 15:23:11,123 INFO The hbdk parameter is not specified, and the submodel will be compiled with the default parameter.

2021-11-29 15:23:11,124 INFO HorizonNN version: 0.11.2

2021-11-29 15:23:11,125 INFO HBDK version: 3.24.4

2021-11-29 15:23:11,126 INFO [Mon Nov 29 15:23:11 2021] Start to parse the onnx model.

2021-11-29 15:23:11,169 INFO ONNX model info:

ONNX IR version: 5

Opset version: 10

Input name: input, [1, 640, 640, 3]

2021-11-29 15:23:11,199 INFO [Mon Nov 29 15:23:11 2021] End to parse the onnx model.

2021-11-29 15:23:11,201 INFO Model input names: [‘input’]

2021-11-29 15:23:11,214 INFO Saving the original float model: ./.hb_check/original_float_model.onnx.

2021-11-29 15:23:11,216 INFO [Mon Nov 29 15:23:11 2021] Start to optimize the model.

./01_check.sh: line 20: 22128 Segmentation fault hb_mapper checker --model-type ${model_type} --model ${caffe_model} --output ${output} --march ${march}

您好,ONNX IR version: 5 可以改成6-
Opset version: 10 改成11

然后再尝试一下,不行的话加我微信Horizon_robot

我试过,10,11,12都试过,最后通过不停调模型结构发现发现是resize算子后直接用add算子,就会出这个错误,resize后随便加个什么层,再弄add就没问题了

但是这不是解决办法呀,中间加个层,模型都要重新训练了?

技术人员正在排查,请耐心等待