PsyProxy Package Options

Overview

Walkthrough asset · Part 1 of 2

Every configurable knob, its default, and the alternatives you can swap in. Click any row to expand. Default value sits on the right as a green pill; the alternatives panel explains when and why to deviate. Image placeholder — options overview Diagram to be supplied. Suggested: a box-and-arrow overview of the three configuration layers (CLI → evaluation config → model config) and how they flow into the validation pipeline shown below. required no default — user must supply green default value None optional, no default A. Command-line interface user-facing entry point Accepts any CSV, re-maps columns to internal names, assigns folds, computes proxy features through a chosen lens, runs the validation pipeline, and writes evaluation artifacts to a user-specified output directory. --input-path required path required Path to input CSV. Any UTF-8 CSV the column map can resolve. Note Missing target values don't abort — the loader drops rows whose target is missing or NA-like. --output-root required path required Destination for inputs/ , systems/<sys>/features/ , systems/<sys>/models/strict|permissive/ , and the rendered benchmark card. Cache behavior If the directory already contains feature matrices, they are reused unless --force-rebuild is set. --text-column required str required Free-text column. NaN→empty string, then coerced to str. Length limits No explicit cap — the encoder truncates to its own maximum sequence length. --target-column required str[+] required One or more outcome columns. Multiple targets spawn one evaluation per target (e.g. --target-column quality difficulty would_take_again ). Pairs with --task-type — provide either one shared type for all targets, or one type per target. --task-type required enum required One of the four supported task types. Drives fold stratification, selection metric, default model family, and the ACE y-transform. Choices regression · linear regression, r² default metric binary · logistic regression, ROC-AUC default multiclass · multinomial logistic regression, accuracy ordinal · proportional-odds ordinal regression, quadratic κ --id-column None Row identifier. If absent or non-uniq