fix: adjust good/bad prompts hack to match tests

This commit is contained in:
Philipp Emanuel Weidmann
2026-09-22 19:26:02 +05:30
parent 598f5f4bd4
commit 033eb4238c
+2 -2
View File
@@ -413,13 +413,13 @@ def run():
# TODO: Introduce a dedicated dataset setting for test prompts. # TODO: Introduce a dedicated dataset setting for test prompts.
good_prompts_dataset = DatasetSpecification( good_prompts_dataset = DatasetSpecification(
dataset="mlabonne/harmless_alpaca", dataset="mlabonne/harmless_alpaca",
split="train[:400]", split="train[:5]",
column="text", column="text",
) )
bad_prompts_dataset = DatasetSpecification( bad_prompts_dataset = DatasetSpecification(
dataset="mlabonne/harmful_behaviors", dataset="mlabonne/harmful_behaviors",
split="train[:400]", split="train[:5]",
column="text", column="text",
) )