Closed tradysun closed 3 years ago
It's not the issue of simple tensorflow serving. But yes, you can run docker container with GPU devices without nvidia-docker
. Please refer to the older document of TensorFlow in https://github.com/tensorflow/tensorflow/tree/r1.8/tensorflow/tools/docker .
This may work for docker image for simple tensorflow serving as well.
$ export CUDA_SO=$(\ls /usr/lib/x86_64-linux-gnu/libcuda.* | xargs -I{} echo '-v {}:{}')
$ export DEVICES=$(\ls /dev/nvidia* | xargs -I{} echo '--device {}:{}')
$ docker run -it -p 8888:8888 $CUDA_SO $DEVICES tensorflow/tensorflow:latest-gpu
My linux system is centos6.3, which only support docker1.7, and does not support nvidia-docker(need docker>=1.13). How to run simple tensorflow serving with GPU?