I’d like to use Triton Inference Server on ppc64le with Tensorflow support.
The only available option seems to build it using CMake. Am I right ?
Configuring with :
cmake -DTRTIS_ENABLE_TENSORRT=ON -DTRTIS_ENABLE_TENSORFLOW=ON -DTRTIS_EXTRA_LIB_PATHS="/home/localuser/tensorflow-lib" …/build
It looks like I need to have a Tensorflow library in order to build libtrtserver.so.
(I will put it in directory “/home/localuser/tensorflow-lib”)
So I need to build Tensorflow on ppc64le too ?
I found a docker (https://hub.docker.com/r/ibmcom/tensorflow-ppc64le) with a Tensorflow version for ppc64le.
but I’m not sure docker is an option for me.
Is there any special option for building Tensorflow on ppc64le for use with Triton Inference Server ?