The repository contains the code (template) used for testing compact LLM models.
- The
QAData.jsonfile contains all 570 questions used to test models from Polish history. - The
judge_belik.ipynbfile contains the code used to evaluate the responses of all tested models in “LLM as judge” mode. - The
check_response_times.ipynbfile contains the code used to determine the normalized response times of the models. - Files with the prefix
0experiment_contain logs from the model tests.