TensorRT-5.1.5.0-ShuffleNet-v2

TensorRT指南

https://blog.csdn.net/hw5226349/column/info/25869

 

caffe版本：

https://github.com/farmingyard/ShuffleNet

测试机器为GTX 1070ti

name: "shufflenet_v2"
input: "data"
input_shape { 
  dim: 1
  dim: 3
  dim: 224
  dim: 224
}

I0715 10:53:01.804968 10581 caffe.cpp:407]     fc1000	backward: 0.0735949 ms.
I0715 10:53:01.804980 10581 caffe.cpp:412] Average Forward pass: 8.58957 ms.
I0715 10:53:01.804991 10581 caffe.cpp:414] Average Backward pass: 8.04892 ms.
I0715 10:53:01.804998 10581 caffe.cpp:416] Average Forward-Backward: 16.6905 ms.
I0715 10:53:01.805004 10581 caffe.cpp:418] Total Time: 3338.1 ms.
I0715 10:53:01.805009 10581 caffe.cpp:419] *** Benchmark ends ***

TensorRT已经支持的激活函数：