Skip to content

Requirements

First off, I would recommend using a Conda virual environment:

conda create --name ML_train
conda activate ML_train

conda install -c conda-forge python ipython jupyterlab 
conda install -c conda-forge scipy pandas numpy networkx scikit-learn
conda install -c conda-forge nbconvert matplotlib seaborn plotly
conda install -c conda-forge streamlit scikit-plot sqlite grip
pip3 install torch torchvision torchaudio
pip3 install mkdocs mkdocs-material
pip3 install mkdocs-pymdownx-material-extras
conda install -c conda-forge r-base r-essentials r-pacman r-psychtools r-lars

We probably don't need all of these packages, but this is the environment I'm working with. You can also try:

    conda env create -f requirements_mac.yml

using the requirements file in the main directory. I have typically found that this only works on the same OS though.

Some notes

  • scikit-learn and torch are for machine learning.
  • sqlite is used to manipulate SQL databases using SQL in Python. (Very useful data structures for industry).
  • mkdocs is a great package for making code documentation websites (Like this one!).
  • nbconvert converts Jupyter notebooks to .py files.
  • matplotlib, seaborn, plotly is for plotting and data visualisation, streamlit is for dashboards.
  • r-base, r-essentials, r-pacman will let us use R. r-pacman is a package manager that will automatically download required packages from CRAN. For deployment, it might be better to download packages using conda. This will work fine for us though.
  • r-psychtools has the package psych which contains useful functions like describe(). On my Mac, I needed to have XCode installed first for this to work.