Training ML Models

Thursday workshop at the Institute of Computing for Climate Science Summer School 2022.

• Slides: training_ml_models.pdf
• Notebook: training_ml_models.ipynb
• Script: training_ml_models.py

Download

git clone https://github.com/handley-lab/2022-cambridge-iccs.git
cd 2022-cambridge-iccs

Installation instructions

You will need to install

• notebook
• matplotlib
• scikit-learn
• tensorflow
• torch

in that order -- you can leave the tensorflow/pytorch installation in the background until exercises 2 and 3 if downloads are slow.

All are pip-installable, which you should do either:

• via pip
• via conda
• via your system installation (apt-get, yum, pacman, etc) (none of these packages are controversial),

The way I do it when not using my arch system packages is with virtual environments:

python -m venv my_venv
source my_venv/bin/activate
pip install notebook matplotlib scikit-learn tensorflow torch
python -m ipykernel install --user --name=my_venv

The final line tells the jupyter notebook where to look for the virtual environment.

Run

Start the jupyter notebook

jupyter-notebook training_ml_models.ipynb

and select the virtual environment we've just installed

FAQ

• How do I convert between notebooks and python scripts? py2nb
• Why would I want to do that? vimteractive