Getting Started¶

Installation¶

git clone https://github.com/rrivera1849/STEB.git
cd STEB
python3 -m venv venv
source venv/bin/activate
pip install -e .

Downloading Datasets¶

Some datasets need to be downloaded before use:

./download_datasets.sh

The script skips datasets that have already been downloaded. Use --purge to force a clean re-download.

Configuration¶

By default, STEB looks for raw datasets in ./raw_datasets relative to the working directory. All paths are configurable via environment variables or a config.ini file.

Environment Variables¶

Variable	Default	Description
`STEB_RESULTS_DIR`	`./results`	Where evaluation results are saved
`STEB_PROCESSED_DATA_DIR`	`~/.local/share/steb/processed_datasets`	Processed dataset cache
`STEB_RAW_DATASETS_DIR`	`./raw_datasets`	Raw downloaded datasets

The HuggingFace load_dataset cache (used for HF-format datasets in the suite) is managed by HuggingFace itself; set HF_DATASETS_CACHE or HF_HOME to relocate it.

Config File¶

These can also be set in a config.ini file (in the current directory or ~/.steb/config.ini):

[Application_Paths]
processed_dataset_dir = /path/to/your/processed_datasets
results_dir = /path/to/your/results
raw_datasets_dir = /path/to/your/raw_datasets