OBSOLETE: now either TFS install via package manager like apt-get or using official dockerfile from here https://github.com/tensorflow/serving/tree/master/tensorflow_serving/tools/docker

tensorflow-serving-cuda-docker

Docker image for tensorflow serving with CUDA which suitable for development purpose

This image is based from nvidia/cuda:8.0-cudnn5-devel-ubuntu14.04 include:

• ubuntu 14.04
• CUDA 8 with development packages
• cuDNN 5 with development packages
• bazel 0.4.4
• tensorflow 1.0.0
• tensorflow serving source code that is ready to build
• other setting: proxy,enviroment variable, etc .

You can using either following dockerfile:

• Dockerfile: no tensorflow serving build happen:
• Dockerfile.GUI extended from above Dockerfile with GUI and VNC server:
  connect to: vnc://:5901 via VNC client. The VNC password is password.
  After build the docker image, run tensorflow serving compliation with the desired setting.

cd /serving/tensorflow
./configure
cd /serving
bazel build -c opt --config=cuda tensorflow_serving/...
bazel test tensorflow_serving/...