From 193b4e0484717b6e149b40e15549b5e757cbd769 Mon Sep 17 00:00:00 2001 From: Jeremy Dohmann Date: Fri, 29 Sep 2023 09:53:43 -0400 Subject: [PATCH] commit --- scripts/eval/eval.py | 14 +++----------- 1 file changed, 3 insertions(+), 11 deletions(-) diff --git a/scripts/eval/eval.py b/scripts/eval/eval.py index f07942ba10..2e94ca99d2 100644 --- a/scripts/eval/eval.py +++ b/scripts/eval/eval.py @@ -314,23 +314,15 @@ def main(cfg: DictConfig): if eval_gauntlet_df is not None and eval_gauntlet_callback is not None: assert composite_scores is not None row = {'model_name': model_cfg['model_name']} - row.update({ - t.name: - composite_scores.get(f'icl/metrics/eval_gauntlet/{t.name}', - None) - for t in eval_gauntlet_callback.categories - }) - row.update({ - 'average': - composite_scores[f'icl/metrics/eval_gauntlet/average'] - }) + row.update({k.split('/')[-1]: v for k,v in composite_scores.items()}) eval_gauntlet_df = pd.concat( [eval_gauntlet_df, pd.DataFrame([row])], ignore_index=True) print(f'Printing gauntlet results for all models') + print( eval_gauntlet_df.sort_values( - 'average', ascending=False).to_markdown(index=False)) + list(eval_gauntlet_df.columns)[-1], ascending=False).to_markdown(index=False)) print(f'Printing complete results for all models') assert models_df is not None print(models_df.to_markdown(index=False))