tensorRT doesn't support PReLu ?

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

I want to use tensorRT to parse caffe model, but the tensorRT only support ReLu layer but not PReLu. So how can I do with it ? Can I define it myself ? if there are some demos to reference ?

2

Hi,

You can implement it with plugin API.
From TensorRT2.1, we release IPlugin API for the user to implement a non-supported layer.

We have several samples to demonstrate plugin API:

Native sample:
/usr/src/tensorrt/samples/samplePlugin/samplePlugin.cpp
/usr/src/tensorrt/samples/sampleCharRNN/sampleCharRNN.cpp
/usr/src/tensorrt/samples/sampleFasterRCNN/sampleFasterRCNN.cpp

TX2 sample:
https://github.com/AastaNV/Face-Recognition

Thanks.