Docker Image for TensorFlow™. This Docker image provides Python, Java, C and Go execution environment.

Provided images :

• MASTER - TensorFlow™ 1.2.1 with Python 2.7 CPU
• TensorFlow™ 1.2.1 with Python 2.7 CPU
• TensorFlow™ 1.2.1 with Python 2.7 GPU
• TensorFlow™ 1.2.1 with Go 1.8.3 CPU
• TensorFlow™ 1.2.1 with Go 1.8.3 GPU

TensorFlow™ is an open source software library for numerical computation using data flow graphs. Nodes in the graph represent mathematical operations, while the graph edges represent the multidimensional data arrays (tensors) communicated between them. The flexible architecture allows you to deploy computation to one or more CPUs or GPUs in a desktop, server, or mobile device with a single API. TensorFlow was originally developed by researchers and engineers working on the Google Brain Team within Google's Machine Intelligence research organization for the purposes of conducting machine learning and deep neural networks research, but the system is general enough to be applicable in a wide variety of other domains as well.

Here some more info about TensorFlow™ : https://www.tensorflow.org/

Here some infromation about TensorFlow™ develoment : https://www.tensorflow.org/get_started/

This docker images has been designed to be a test, development, integration, production environment for TensorFlow™. No warranties for production use.

Here some information :

Volumes : /root/tf-app, /root/.tensoboard/

/root/tf-app :

Folder to install sources.

/root/.tensoboard :

Folder collecting logs and event logs.

Ports:

6006, 88888, 22

6006 :

TensoBoard™ WebUI Port

8888 :

IPython Jupyter WebUI Port

22 :

SSH port (ssh public key will be printed in container logs)

Here TensorFlow® environment variables :

• JUPYTHER_TOKEN : Jupyter access token (default: "7e7f9117ae5b96a8e69126ccb70841ec2911a051c6bb4ba7")

Here some auto-install source form remote source, environment variables :

• TARGZ_ROOT_SSH_KEYS_URL : URL to download tar gzipped root user ssh keys (default: "")
• TARGZ_USER_SSH_KEYS_URL : URL to download tar gzipped custom defined (jupyter) ssh keys (default: "")
• TARGZ_SOURCE_URL : URL to download tar gzipped source code (default: "")
• GIT_URL : Git repository URL (default: "")
• GIT_BRANCH : Git repository desired branch (default: "master")
• GIT_USER : Git repository user (default: "")
• GIT_EMAIL : Git repository email (default: "")

Here a sample command to run TensorFlow™ container:

docker run -d -v my/app/dir:/root/tf-app -p 8888:8888 -p 6006:6006 --name my-tensiorflow hellgate75/tensiorflow:latest

You can run container with -bash argument for an on-flight execution and destroy, as follow :

docker run --rm -v my/app/dir:/root/tf-app -p 8888:8888 -p 6006:6006 --name my-tensiorflow hellgate75/tensiorflow:latest -bash my-command my-argument-1 ...  my-argument-n

NOTE:

For GPU docker container versions, please use nvidia-docker available at :

https://github.com/NVIDIA/nvidia-docker/wiki/Installation

You can enforce nvidia drivers and devices running :

nvidia-docker run [-d | --rm] --privileged  -v my/app/dir:/root/tf-app -p 8888:8888 -p 6006:6006 --name my-tensiorflow hellgate75/tensiorflow:latest ....

TensorFlow™ TensoBoard event log folder is : /root/.tensoboard, please refer to this folder or use environment variable TENSORFLOW_LOG_FOLDER to set-up code development reference to log event folder.

In order to access to TensorFlow™ shell :

    docker exec -it my-tensiorflow bash

Then, into docker container, type :

    python /root/tests/test.py

In order to test TensorFlow™ TensoBoard, open in your browser :

http://{ host name | ip address | localhost }:6006
eg.:
http://localhost:6006

In order to test TensorFlow™ Jupyter Notebook Board (for testing and modify source), open in your browser :

http://{ host name | ipaddress | localhost }:8888/token={ configured token: JUPYTER_TOKEN }
eg.:
http://localhost:8888?token=7e7f9117ae5b96a8e69126ccb70841ec2911a051c6bb4ba7

LGPL 3