I wanted to play around with the docker version of tensorflow while I’m trying to fix build breaks on the gpu-accelerated Windows TS deployment I’m playing with.
There’s already a TS docker image. I needed to get it and modify it. Here’s the steps I did to do that.
Have a Windows Machine running Docker, either via VirtualBox or Hyper-V. You’ll need to know how to set a port forwarding rule to the default docker VM.
- pull the image. “docker pull gcr.io/tensorflow/tensorflow”
- run the image “docker run -it -p 8888:8888 gcr.io/tensorflow/tensorflow”
- Exec a shell:
- “docker ps” ( find the container ID )
- “docker exec -it [IMAGE] bash
- install scikit-learn via pip in the image: pip install scikit-learn
- exit the bash shell
- Create a port forward rule from localhost:[PORT] to [default:8888].
- shutdown the jupyter notebook running by default in the TS image.
- docker commit [IMAGE] docker-local
You can now run the image you created with:
docker run -it -p 8888:8888 tensorflow-local
This gives you a jupyter notebook server with TS and scikit-learn as a docker machine.
Now, if only nvidea-docker would work on Windows. A man can dream…