CODEWITHSUNDEEP
pythondockertensorflowdocker-composejupyter-notebook
SumNeuron
SumNeuron

Reputation: 5198

Tensorflow 1.14+: docker image doesn't work as expected with docker-compose

I am trying to test my setup using docker and tensorflow. I am using the official Tensorflow image tensorflow/tensorflow:1.15.0rc2-gpu-py3

My project has the minimum structure:

project/
    Dockerfile
    docker-compose.yml
    jupyter/
        README.md

I have the following Dockerfile:

# from official image
FROM tensorflow/tensorflow:1.15.0rc2-gpu-py3-jupyter
# add my notebooks so they are a part of the container
ADD ./jupyter /tf/notebooks


# copy-paste from tf github dockerfile in attempt to troubleshoot
# https://github.com/tensorflow/tensorflow/tree/master/tensorflow/tools/dockerfiles/dockerfiles/gpu-jupyter.Dockerfile
WORKDIR /tf
RUN which jupyter
CMD ["jupyter-notebook --notebook-dir=/tf/notebooks --ip 0.0.0.0 --no-browser --allow-root"]

and the docker-compose.yml

version: '3'

services:
  tf:
    image: tensorflow/tensorflow:1.15.0rc2-gpu-py3-jupyter

    # mount host system volume to save updates from container
    volumes:
      - jupyter:/tf/notebooks

    ports:
      - '8888:8888'

    # added as part of troubleshooting
    build:
      context: .
      dockerfile: Dockerfile


volumes:
  jupyter:

running docker-compose build and docker-compose up succeeds (if the CMD in the Dockerfile is commented out), but just exits. From the docker hub repository, I thought adding the volume would auto-start a notebook.

Trying to run jupyter-notebook or jupyter notebook fails.

Thoughts on how to correct?

Upvotes: 1

Views: 1147

Answers (2)

Ali Ganjbakhsh
Ali Ganjbakhsh

Reputation: 801

try this

RUN pip3 install nvidia-tensorflow

this will install tf 1.15

Upvotes: 0

ikolomiyets
ikolomiyets

Reputation: 146

If you want to create a custom image from the official one adding the notebook directory then the image property in the docker-compose should be the name of the your local image not the tensorflow/tensorflow:1.15.0rc2-gpu-py3-jupyter. All you need is in this case a following Dockerfile:

FROM tensorflow/tensorflow:1.15.0rc2-gpu-py3-jupyter
ADD ./jupyter /tf/notebooks

In this case the docker-compose.yaml file should look like the following:

version: '3'
services:
  tf:
    image: tensorflow

    # mount host system volume to save updates from container
    volumes:
      - jupyter:/tf/notebooks

    ports:
      - '8888:8888'

    # added as part of troubleshooting
    build:
      context: .
      dockerfile: Dockerfile


volumes:
  jupyter:

Note, that the image is tensorflow.

However, there is really no need to use the custom Dockerfile. Just use the following docker-compose.yaml file:

version: '3'
services:
  tf:
    image: tensorflow/tensorflow:1.15.0rc2-gpu-py3-jupyter

    # mount host system volume to save updates from container
    volumes:
      - ./jupyter:/tf/notebooks:Z

    ports:
      - '8888:8888'

It will directly map your local jupyter directory to the container and will use an official image without modification.

Note though, it might not work as expected on Windows due to issues with mapping of the host directories.

Upvotes: 1

Related Questions