.. _installation-instructions:

=======================
Installation
=======================

Quick Install
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

PiML currently supports various os platforms and python versions, including [py37, py38, py39, py310] * [Win, Linux, and Macos]. Starting from PiML-0.5.1, Mac M1 with py38, py39, py310 is supported.  

You can install piml directly on these environments by 

.. prompt:: bash $

    pip install piml

For Mac M1 users, cmake is needed before installing piml. The following command can be used to install cmake.

.. prompt:: bash $
	
	conda install cmake


Dependencies
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

PiML requires the following packages:

======================= ===================
Dependency              Minimum Version    
======================= ===================
ipykernel
ipywidgets              7.7.0
joblib                  1.1.0
ipython                 7.12.0
numpy                   1.21.4 (<1.24.0)
scipy                   1.5.3
pandas                  1.3.5 (<2.0.0)
matplotlib              3.1.2 (<3.8.0)
seaborn                 0.11.2
xlrd                    1.2.0
scikit-learn            0.24.2 (<1.4.0)
xgboost                 1.4.2
statsmodels             0.12.2
lime                    0.2.0.1
shap                    0.39.0
torch                   1.11.0
pygam                   0.8.0
natsort                 8.2.0
psutil                  5.9.0
dill                    0.3.6
optbinning              0.17.3
ortools                 9.4 (<9.5.0)
packaging               20.5
networkx                2.6.3
momentchi2              0.1.8
jupyter_client          (<=7.4.9)
======================= ===================

The following packages are optional

======================= ===================
Dependency              Minimum Version    
======================= ===================
dcor                    0.5.3
solas-ai                0.2.4
h2o                     3.32.0.0
pyspark                 2.3.0
======================= ===================


Troubleshooting
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Could not find a version that satisfies the requirement PiML
""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
If you got the following error message:

.. prompt:: bash $

	ERROR: Could not find a version that satisfies the requirement PiML
	ERROR: No matching distribution found for PiML

One possibility is that your Python version is not one of the [py37, py38, py39, py310].

If your Python version is correct, you may need to upgrade pip first and then reinstall piml.

.. prompt:: bash $

    pip install pip -U 
    pip install piml

Cannot uninstall "llvmlite".
""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
Try the following command.

.. prompt:: bash $

    pip install llvmlite --ignore-installed

Library "libxgboost.so" not loaded
""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
This is an error related to the xgboost package.
You may reinstall it using conda.

.. prompt:: bash $

    pip uninstall xgboost
    conda install -c conda-forge py-xgboost

Cannot import PiML on Colab
""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
After running pip install piml on Colab, you may get the following error message:

.. prompt:: bash $

    RuntimeError Traceback (most recent call last)
    RuntimeError: module compiled against API version 0x10 but this version of numpy is 0xf

To fix this, you may need to click on "Restart runtime" and then import piml directly:

.. prompt:: bash $

	from piml import Experiment
	exp = Experiment()