Jetpack 4.2 : Cannot deserialize plugin FlattenConcat_TRT

Hi Supporter,

I’m using Jetpack 4.2.2 with TensorRT 5.1.6.

I created successfully engine file from UFF file by using python script and link to this library libflattenconcat.so
ctypes.CDLL("/usr/local/lib/libflattenconcat.so")

But when load the engine file using c++ code, I linked this libflattenconcat.so to application by using Cmake. However I can’t deserialize the engine because lack of FlattenConcat_TRT plugin.

Plugin Creator registration succeeded - GridAnchor_TRT
Plugin Creator registration succeeded - NMS_TRT
Plugin Creator registration succeeded - Reorg_TRT
Plugin Creator registration succeeded - Region_TRT
Plugin Creator registration succeeded - Clip_TRT
Plugin Creator registration succeeded - LReLU_TRT
Plugin Creator registration succeeded - PriorBox_TRT
Plugin Creator registration succeeded - Normalize_TRT
Plugin Creator registration succeeded - RPROI_TRT
Plugin Creator registration succeeded - BatchedNMS_TRT
getPluginCreator could not find plugin FlattenConcat_TRT version 1 namespace 
Cannot deserialize plugin FlattenConcat_TRT
getPluginCreator could not find plugin FlattenConcat_TRT version 1 namespace 
Cannot deserialize plugin FlattenConcat_TRT

Did you meet this problem before and if yes, could you share me how do you can overcome this issue ?

Many thank in advance !

Regards,
Khang

Is this a plugin you write? did you add
REGISTER_TENSORRT_PLUGIN(luginCreator); in your code? or as it said, there are something wrong with plugin namespace

Hi @zerollzeng,

No, this is not a plugin I write, I took from TensorRT source code, branch 5.1.

I’m too naive to believe that only link to libflattenconcat.so like python, I can register this plugin.

I got the answer from “/usr/src/tensorrt/samples/sampleUffSSD” and how to register the plugin.

Thank you for your answer.

Hi,

You can also refer to the below sample which uses FlattenConcat plugin:
https://github.com/NVIDIA/TensorRT/tree/07ed9b57b1ff7c24664388e5564b17f7ce2873e5/samples/opensource/sampleUffSSD

Thanks