Skip to content

Latest commit

 

History

History
96 lines (53 loc) · 4.68 KB

piov097aml.md

File metadata and controls

96 lines (53 loc) · 4.68 KB

Installing PredictionIO-0.9.7-aml

PredictionIO-0.9.7-aml is that last stable version that uses the Java namespace io.prediction (look in the templates build.sbt file, it will reference io.prediction instead of org.apache.predictionio), if your template uses this you may want to install this version of PIO. We recommend that everyone move to the newer Apache version installed here.

If you are using The Universal Recommender v0.4.2, for example and for some reason can't upgrade to the Apache compatible Universal Recommender v0.5.0 follow these steps.

IMPORTANT NOTES:

  1. Do not install as the "root" user. All guides assume that you have a user account that has sudo permission. You will have trouble if you install as "root". If your machine is clean do the following:

    1.1 Create user for PredictionIO aml in each server

    adduser aml # Give it some password

    1.2 Give the aml user sudoers permissions and login to the new user. This setup assumes the aml user as the owner of all services including Spark and Hadoop (HDFS). Do not install or run as root!

    sudo nano /etc/sudoers.d/sudo-group

    Add this line to the file

    # Members of the sudo group may gain root privileges
# with no password (somewhat controversial)
%sudo  ALL=(ALL) NOPASSWD:ALL

    Then save and add aml user to sudoers

    sudo usermod -a -G sudo aml
sudo su aml # or exit and login as the aml user
cd ~
sudo service sudo restart # just to be sure permission are all active

  2. You need to build pio on your machine since this sets up class caches in ~/.ivy2/... and before you build or download you should erase the classes in that cache with rm -r ~/.ivy2

Upgrade

If you already have some version of PIO you should upgrade.

Move old PIO

  1. mv /path/to/pio pio-old move the directory containing the old version of PredictionIO, keep it for the configuration.
  2. rm -r ~/.ivy2 This is required and will remove the old version of the local cache of classes created when building PredictionIO and templates.

Download ActionML's Release of PredictionIO

ActionML has enhanced the original version of PredcitionIO v0.9.5 and has been enhancing it through several new versions. Our templates rely on these enhancements so use these installation instructions. to get the right version—do not rely on docs.prediciton.io

  1. git clone https://github.com/actionml/PredictionIO.git pio-aml clone to some directory. Make sure you are on the master branch or use the tag for the desired release version. git checkout master is usually what you want.

Build PredictionIO

You must build PredictionIO from source to get needed classes installed in your local cache or the Universal Recommender will not build, you will get error in pio build. This is pretty easy:

  1. cd /path/to/pio-aml
  2. ./make-distribution

Configuration

If you have installed a recent version of PIO in the past, copy the configuration for it.

  1. cp old-pio/conf /path/to/pio-aml

Make sure to configure all components using one of the setup guides here.

To test your installation run pio status to make sure pio is working. Also check to make sure HDFS and and Spark are running correctly since pio status does not check the running status of those services.

Install Fresh

For a first-time install on a single machine. NOTE: do not install as "root"! Create a sudoer user as described above.

Install From a Script

For a completely fresh new install, do not use the script on PredictionIO's docs site do the following:

  1. bash -c "$(curl -s https://raw.githubusercontent.com/actionml/PredictionIO/master/bin/install.sh)"

This will create a vendors subdirectory with needed services installed there. It will also trigger a build of PIO from source so make sure you put it in a place where you have permission to read/write.

Note: this is only for a single machine developer setup and is not advised for production.

Manual Installation Guides

{{> piosetupguides}}

Templates

  • The Universal Recommender

    Use the installation described here. The primary exception to the quickstart is that you are using a version with the io.prediction namespace, so the UR v0.4.2 or below not the UR v0.5.0 or above, these will have build errors.

  • Build Any Template

    Building a template that uses the io.prediction namespace (non-Apache PredictionIO) with the install instructions above is the same as normal:

    git clone https://github-repo.git/path/to/template /path/to/template/directory
cd /path/to/template/directory
pio build