can tensorRT3 support tf.argmax operation

can tensorRT3 support tf.argmax operation

Hi,

TensorRT 3 doesn’t support argmax layer.
If your argmax function is at the end of the networks, you can implement it with C++/python directly.

Thanks.

hi AastaLLL
How to implement argmax layer with C++, is there any example code or some refences?
thank you~

hi AastaLLL
How to implement argmax layer with C++, is there any example code or some refences?
thank you~

Hi,

You can implement a custom layer with plugin API, but this is only for the Caffe model.
We don’t provide an interface for a Tensorflow/UFF user to set their plugin implementation yet.

If you want to use argmax layer, please switch to Caffe framework and write it with plugin API.

Here is our sample for plugin interface:
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.