GitHub

Deborah.DeborahDocument.DeborahDocumentRunner

Deborah.DeborahDocument.DeborahDocumentRunner.run_DeborahDocument — Function
run_DeborahDocument(
  toml_path::String, 
  jobid::Union{Nothing,String}=nothing
) -> Nothing

Run the full Deborah.DeborahCore document-generation pipeline for a single ensemble.

Overview

  • Reads a TOML config, loads/extends summary data, saves them into JLD2 format.

Config (TOML) expectations

  • The file at toml_path must contain, at minimum, the following sections/keys:

    [data]

    • location::String : project root (e.g., "./nf4_clover_wilson_finiteT")
    • ensemble::String : ensemble name (e.g., "L8T4b1.60k13570")
    • analysis_header::String : analysis prefix (e.g., "analysis")
    • X::Vector{String} : input observable file names (e.g., ["plaq.dat","rect.dat"])
    • Y::String : target observable file name (e.g., "pbp.dat")
    • model::String : learner/model tag (e.g., "LightGBM")
    • labels::Vector{String} : label indices as strings (later parsed to Int)
    • trains::Vector{String} : train indices as strings (later parsed to Int)
    • use_abbreviation::Bool : whether to abbreviate observable names in outputs

    [bootstrap]

    • ranseed::Int, N_bs::Int, blk_size::Int

    [jackknife]

    • bin_size::Int

    [abbreviation]

    • Dict{String,String} mapping raw filenames to short tags (e.g., "pbp.dat" => "TrM1", "plaq.dat" => "Plaq", ...)

Arguments

  • toml_path::String : Path to the TOML configuration file.
  • jobid::Union{Nothing,String} : Optional job identifier appended to model suffixes for logging/auditing.

Outputs

  • Results snapshot files: results_<overall_name>.jld2 saved under the analysis path and copied to HERE.

Workflow

  1. Parse configuration from toml_path.
  2. Build name strings:
    • analysis_ensemble = "<analysis_header>_<ensemble>"
    • overall_name = "<ensemble>_<learning>"
    • cumulant_name = "<analysis_header>_<overall_name>"
    • where learning is either abbreviated (XY_code) or full (X_Y) with model(+jobid) suffix.
  3. Load and extend summary data via Deborah.Rebekah.SummaryLoader.load_summary.
  4. Save results to JLD2 (results_<overall_name>.jld2) and copy to HERE.

Notes

source