epibench fetch

Running epibench fetch <challenge-id> --output-path ".../..."via the command line will fetch all of a challenge's data files to your machine (at the output directory specified by the --output-path flag). After a successful run, you will find the following:

• a directory titled <challenge-id>
  • agent.md: a file to inform an agent where to find instructions associated with the challenge
  • instruction.md: a file with a description of how to properly utilize the fetched data for forecasting, including the columns your model output should contain
  • <challenge-id>.json: the JSON file that internally defines the challenge
  • task_list.csv: a CSV file with two columns; the reference date and the coresponding relative path to the ground truth vintaged to that date
  • gt/: a folder that contains the ground truth data
    • each ground truth data file is nested in gt/YYYY-MM-DD/YYYYMMDD.csv, where YYYY-MM-DD is a given reference date

example usage:

epibench fetch epb_flu_inchosp_2024-2025_dev --output-path "/Users/user/Desktop"