Skip to content
NFL Spread Predictor Pipeline

NFL Spread Predictor Pipeline #37

Sign in to view logs

NFL Spread Predictor Pipeline

NFL Spread Predictor Pipeline #37

Workflow file for this run

.github/workflows/train-spread-predictor.yaml at ce7e843
name: NFL Spread Predictor Pipeline
on:
# schedule:
# - cron: '0 0 * * 2'
workflow_dispatch:
jobs:
build:
runs-on: ubuntu-latest
permissions:
contents: write
steps:
- name: Checkout repository
uses: actions/checkout@v4
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: 3.12
- name: Install Poetry
uses: snok/install-poetry@v1
with:
version: 1.7.1
- name: Install dependencies with Poetry
run: poetry install
- name: Download data
run: poetry run python ./nfl_analytics/main.py --download
# Necessary for train step
# https://github.com/actions/runner-images/discussions/7188#discussioncomment-6750749
# https://stackoverflow.com/questions/71590851/r8-is-causing-gradle-daemon-to-vanish-on-github-hosted-action-runner/76921482#76921482
- name: Increase swapfile
run: |
df -h
sudo swapoff -a
sudo fallocate -l 12G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
sudo swapon --show
# Seperated from Download data to make debugging easier
- name: Train Model
run: poetry run python ./nfl_analytics/main.py --train
- name: Create release
id: create_release
uses: softprops/action-gh-release@v1
with:
files: ./nfl_analytics/assets/*
tag_name: spread-predictor
draft: true
body: |
This release includes the trained model, scaler, and compressed CSV file needed for predictions:
The machine learning model saved using joblib.
- **trained_model-[timstamp].joblib:** The scaler pickled with joblib for scaling matchup inputs.
- **trained_scaler-[timstamp].joblib:** The scaler pickled with joblib for scaling matchup inputs.
- **running_average-[timstamp].csv.gz:** Running averages used to form matchup inputs
To make predictions, use these with the main.py --predict command on asset sets with matching timestamps.