error in testing resnet50

[ghimiredhikura]

I got following error while testing tengine resnet50 in my ubuntu machine. I kept default setting in etc/config file. Any idea please?

./RESNET images/air.jpg ../../../models/resnet50/
Successfully load plugin : ../../build/operator/liboperator.so
OPERATOR PLUGIN INITED
Successfully load plugin : ../../build/serializer/libserializer.so
SERIALIZER PLUGIN INITED
Successfully load plugin : ../../build/executor/libexecutor.so
Register reference's operators
EXECUTOR PLUGIN INITED
Successfully load plugin : ../../build/driver/libdriver.so
DEV ENGINE PLUGIN INITED
Driver: RK3399 probed 5 devices
tensor: prob created by node: prob is not consumed
add the node: prob into output list
load resnet50 model done!
failed to set node ops for node: res2a op: Eltwise
graph_task.cpp:103 failed to Prerun task on dev executor: cpu.rk3399.a72.all
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
graph_task.cpp:193 bad status: CREATED
Segmentation fault (core dumped)

[lyuchuny3]

I can run it successfully with default etc/config. Could you please verify your resnet50 model. You can find all the model files required by the examples in [Tengine model zoo)[https://pan.baidu.com/share/init?surl=LXZ8vOdyOo50IXS0CUPp8g] ,(psw: 57vb)

[ghimiredhikura]

well, thank you for your response. I will test with the models in tengine model zoo.
I wanted to use tengine to test caffe SFD model for face detection. But it was unsuccessful and gives same error.
https://github.com/sfzhang15/SFD
I would love to hear your experience if you test above model.
Thanks.

[lyuchuny3]

It seems that SFD is based on SSD. Tengine will release a new version to support ssd operators in these two weeks. You can try at that time.

[ghimiredhikura]

Yes SFD is based on SSD. That is great news, waiting to test SFD in Tengine. Thanks.

[derek-zr]

@ghimiredhikura I also try SFD on tengine. The model was tranformed to tmfile successfully. But when i run it, i get the error: infer shaped for node: node_102 op: Eltwise failed. Have you test SFD on tengine successfully?