Features

Cordage can be used to manage experimental configurations and can aide in managing experimental outputs.

Experimental Configuration

Cordage helps in providing a configuration to a main function. This set of features is the default and (at the moment) cannot be disabled.

CLI and Config Reader

When calling cordage.run with a function, a command line interface is build. This enables you to set the configurations specified in a config dataclass to be set via the command line or pass a configuration files which is automatically read into an respective config object.

Series of Experiments / Multiruns

By specifying the __series__ key in a passed config file, the experiment can be repeated multiple time with varying configuration. __series__ can either contain a list of changes to apply to the base configuration or a (nested dictionary) with mutliple lists (in this case each combination of values is executed once).

Examples

The following will execute 3 repetitions:

{
    "__series__": [
        {"a": 1},
        {"a": 2},
        {"a": 3}
    ]
}

This second example will execute the experiment 4 times:

{
    "__series__": {
        "a": [1, 2],
        "b": [1, 2]
    }
}

Output directory (optional)

Cordage creates an output directory for each experiment based on the configured pattern and stores the If configured (see below), the log-files are automatically saved here.