Interpretable AI
Interpretable AI (IAI) is a suite of machine learning tools that aim for full transparency and understandability by humans. IAI provides a Python interface for their algorithms and can only be used under license. IAI provides a free license for academic use which can be accessed on Alliance systems. For licensing issues, please contact technical support.
Installation on Alliance clusters¶
-
Go to the IAI download page and download the latest release of IAI. Be sure to select "Linux" and the correct version of Julia (1.11.3 in these instructions).
-
Upload the files from Step 1 to the cluster of your choice. Unzip the file, note the location of the files. There should be a
sys.soandArtifacts.tomlfile in the directory. In this example, we'll say/path/to/sysimage/ -
Load requisite modules:
-
Make a Python virtual environment and install IAI:
-
Install IAI in Julia and build PyCall to work in your virtual environment. Make sure your virtual environment is active before launching Julia:
julia> import Pkg; Pkg.add(url="https://github.com/InterpretableAI/IAISystemImages.jl") julia> import IAISystemImages julia> IAISystemImages.install_artifacts("/path/to/sysimage/Artifacts.toml") julia> using Pkg, PyCall julia> ENV["PYTHON"] = joinpath(ENV["VIRTUAL_ENV"], "bin", "python") julia> Pkg.build("PyCall") julia> exit() -
Finish IAI setup in Python:
-
Import IAI for use in your Python scripts
IMPORTANT NOTE¶
Important Note
Any Python script using the interpretableai package has to include setting up Julia to run with the IAI system image. This means the following code must be included before importing IAI in your Python scripts.
It is also a good idea to disable automatic updating in your job script: ```bash export IAI_DISABLE_UPDATE_CHECK=true