lewes6369

It should not be the bottleneck for generating engine.You can save engine only for the first time, and latter you can load from engine file.

I am not sure your issues. As @zerollzeng said, the engine is not generic for different architecture cards. Maybe just try to set CUDA_VISIBLE_DEVICES to the value which graphic card...

I'm sorry your code is something hard to read. And I saw some efficiency problems or maybe some written mistakes on your code. As I see your bug may be...

Hi @JoelYang-github what is your TensorRT version? I only tried in 5.0.2.6/4.0.1.6.

Hi,@JoelYang-github ,which abstract function are not implemented? It seems that I have already override the virtual funcs inherited from the IPluginExt. Can you show me the funcs ,Thanks.

HI,@SyGoing. For PRelu, you can just call createPReLUPlugin in PluginFactory.h checked by the layer name. It is implemented by tensorRT, but you need to load the alpha param.

Thanks for your sharing.In my code I just used as LeakyRelu. Could you share the url of the work mentioned above ? What is the network tested?

Thanks. I will have a look :)