Public | Automated Build

Last pushed: a year ago
Short Description
ML library only support with GPU support
Full Description

RadioML Docker Image

A docker image provided by which provides many of the primitives needed for radio machine learning experimentation.

Docker Image Contents

  • Base: Ubuntu 16.04 Xenial Xerus
  • Remote: ssh-server, x2go server + xfce4, ipython/jupyter notebook
  • Misc: screen, tmux, vim, emacs, git, meld
  • DL: Theano, TensorFlow, Keras, OpenAI Gym, KeRLym
  • ML: Scikit-learn, OpenCV, PyOpenPNL, Pandas
  • SDR: GNU Radio + several useful out-of-tree gr-modules

Quickstart: Downloading and Running Pre-Built Docker-Hub Images

The easiest way to use this image is to pull a pre-built version directly from docker hub

# Get the docker image from the whale cloud
docker pull radioml/full

# Run it (or use various running recipes below)
docker run -i -t radioml/full /bin/bash

Building the Container

Please note: your docker image max size must be >10GB for this build, please see Notes section.

git clone rml
cd rml && sudo docker build -t radioml/radioml . 

This will take a while to build, so find something to do for an hour

Running the Container

To launch in foreground terminal

docker run -i -t radioml/radioml /bin/bash

To launch in background with ssh up (needed before x2go)

docker run -d -P --name test_rml radioml/radioml
docker port test_rml 22
docker port test_rml 8888

Connect with CLI

sudo docker exec -i -t test_rml /bin/bash


ssh root@`docker port test_rml 22`
# use password radioml

Connect with x2go (good way to run GRC)

docker port test_rml 22
# set ssh ip and port from docker, login with root/radioml, use xfce as window manager

Connect with iPython Notebook (good way to run python experiments)

sudo docker exec -i -t test_rml /bin/bash
cd /root/src/notebooks/
ipython notebook

now open http://docker_ip:8888 in the host browser

Using the Image

Launching GNU Radio Companion


Running Keras Examples

cd /root/src/keras/examples

Running KeRLym Examples

cd /root/src/kerlym/examples
KERAS_BACKEND='tensorflow' ./

Running PyOpenPNL Examples

cd /root/src/PyOpenPNL/examples


  • GPU Support: To build with GPU support for use with nvidia-docker, use dockerRML/full-GPU/Dockerfile
  • Image Size: Current sizes are Full: 10.3GB, Full-GPU: 10.5GB, MinimalML: 4.0GB, MinimalSDR: 8.3GB
  • Build Time: Building Full on an 8 core i7-5930K within an RHEL 7.2 KVM instance on a non-SSD raid takes just over 2 hours, YMMV
  • Docker BaseSize: default docker basesize is 10GB, you must increase this to 20GB or 50GB by adding ' --storage-opt dm.basesize=50G ' to DOCKER_OPTS in /etc/default/docker or /etc/sysconfig/docker and restarting the docker daeming (This must be done before starting the build)
Docker Pull Command
Source Repository