This docker image contains Tensorflow Serving (TFS) 1.8 with GPU support and a ready to run inceptionv3 model.
The resulting image can be used in conjunction with the helm chart from Tensorflow Inception Model Chart .
With plain docker you can run via:
docker run -i --expose=9000 quay.io/thomasjungblut/tfs-inception tensorflow_model_server --port=9000 --model_name=inception --model_base_path=/serving/inception-export
For GPU support you can either use nvidia-docker
or the nvidia
runtime.