Run the evaluator locally (check the requirements.txt). Evaluator expects the file with predictions in <predictions-dir>/answers.jsonl and the file with ground truth in <truth-dir>/truth.jsonl. The output will be written to <output-dir>.

~$ python3 evaluator/evaluator.py -a <predictions-dir> -i <truth-dir> -o <output-dir>

Build the container for the evaluator:

$ docker build -t <dockerhub-user>/pan23-authorship-verification:latest -f Dockerfile .

tbd.