Skip to content

mattkinsey/bucky

Repository files navigation

Bucky Model

Documentation Status black-flake8-isort-hooks CodeFactor Interrogate Coverage Status

Documentation

Developer Guide

The Bucky model is a spatial SEIR model for simulating COVID-19 at the county level.

Getting Started

Requirements

The Bucky model currently supports Linux and OSX and includes GPU support for accelerated modeling and processing.

  • git must be installed and in your PATH.
  • GPU support requires a cupy-compatible CUDA installation. See the CuPy docs for details.

Installation

Standard installation:

pip install bucky-covid

Choose a working directory

Bucky will produce multiple folders for historical data and outputs. It's recommended to put these in their own directory, for example ~/bucky

BUCKY_DIR=~/bucky
mkdir $BUCKY_DIR
cd $BUCKY_DIR

Configuration

The default configuration for bucky is located here. Currently, you can locally modify these options by creating a bucky.yml in BUCKY_DIR that will override any of the default options specified in it.

TODO this is WIP and does not work yet:

To use a customized configuration you first need to make a local copy of the bucky configuration. In your working directory:

bucky cfg install-local

Download Input Data

To download the required input data to the data_dir specified in the configuration files (default is $(pwd)/data:

bucky data sync

Running the Model

To run the model with default settings and produce standard outputs.

bucky run

Equivalently, one can the following command (to provide cli configuration to each part of the process)

bucky run model
bucky run postprocess
bucky viz plot

CLI options

Each bucky command has options that can be detailed with the --help flag. e.g.

$ bucky run model --help

Usage: bucky run model [OPTIONS]

  `bucky run model`, run the model itself, dumping raw monte
  carlo output to raw_output_dir.

Options:
  -d INTEGER         Number of days to project forward
                     [default: 30]
  -s INTEGER         Global PRNG seed  [default: 42]
  -n INTEGER         Number of Monte Carlo iterations  [default:
                     100]
  --runid TEXT       UUID name of current run  [default:
                     2022-06-04__08_00_03]
  --start-date TEXT  Start date for the simulation. (YYYY-MM-DD)
  --help             Show this message and exit.

Further CLI documentation is available in the documentation.