These notebooks accompany several talks and workshops developed by Will Benton and Sophie Watson.
- Our workshop "From Statistics to Serverless: Intelligent Applications on OpenShift 4" was delivered at Red Hat Summit and IBM Think in 2020. Slides are available as a PDF or as a movie. The rest of the instructions in this README will cover getting the basic application running.
- Our GTC 2021 talk "Fighting Fraud With One App In Many Ways: GPU-Accelerated End MLOps on Kubernetes" built up a similar fraud-detection application in two versions using RAPIDS.ai. Notebooks from that talk are on this branch.
In order to build and run a model service, you'll need an OpenShift cluster, but you can experiment with the notebooks on your own time. Here's how:
Use binder. (We don't recommend this if you'll be running the tutorial over conference wifi, but it requires almost no setup and can run from a computer that only has a browser.)
If you want to experiment with the data generator, you'll want to use your own computer.
- Make sure you have Python 3.7 installed, installing it if necessary
- If you have a favorite package manager, use that
- if not, python.org has binaries for many platforms
- Make sure you have
git
installed, installing it if necessary- If you have a favorite package manager, use that
- if not, git-scm.com has binaries for many platforms (you won't need a GUI)
- Install pipenv
- on a Mac, the easiest way is probably
brew install pipenv
- on a Fedora Linux machine, the easiest way is probably
dnf install pipenv
- on Windows, if you have Python installed already, the easiest way is probably to use
pip
- on a Mac, the easiest way is probably
- Clone this repository:
git clone https://github.com/willb/fraud-notebooks/
- tip: if you don't have
git
installed, you can also download an archive of this repository
- tip: if you don't have
- Change to this repository's directory:
cd fraud-notebooks
- Install the dependencies:
pipenv install
- Run the notebooks:
pipenv run jupyter notebook