Installation

There are two ways to use AlphaD3M: 1) via Docker/Singularity containers, and 2) via PyPI installation.

Docker/Singularity containers

This version works with Python 3.6 through 3.8 in Linux and Mac. It supports all the ML tasks and data types mentioned here. You need to have Docker or Singularity installed on your operating system. In this version, a container runs in the background, but AlphaD3M continues to be used as a standard Python library.

To install, run:

$ pip install alphad3m-containers

Once the installation is completed, you need to pull manually the Docker image of AlphaD3M.

For Docker:

$ docker pull registry.gitlab.com/vida-nyu/d3m/alphad3m:latest

or for Singularity:

$ singularity pull docker://registry.gitlab.com/vida-nyu/d3m/alphad3m:latest

PyPI

Currently, this version has supports all the ML tasks and data types mentioned here. (using a limited set of primitives). This package works with Python 3.8 in Linux and Mac. Installation will require a version of pip >= 20.3 to leverage the improved dependency resolver, as lower versions may raise dependency conflicts. You might need GCC or other C/C++ compilers to install packages like NumPy, which uses C extensions.

To install, run these commands:

$ pip install alphad3m
$ pip install d3m-common-primitives d3m-sklearn-wrap dsbox-corex dsbox-primitives sri-d3m distil-primitives rpi-d3m-primitives kf-d3m-primitives autonbox lupi-mfa d3m-jhu-primitives fastlvm d3m-esrnn d3m-nbeats --no-binary pmdarima

The second command installs the primitives available on PyPI.

⚠️ WARNING ⚠️

On non-Linux platforms:

• You will need swig to compile pyrfr. You can obtain swig from homebrew, anaconda, or the swig website.

• To install fastlvm primitives on Mac, run:

  CFLAGS=-mmacosx-version-min=10.12 CXXFLAGS=-mmacosx-version-min=10.12 python -m pip install fastlvm