feat: LLM-based Evaluators have raise_on_Failure=False by default (#16

) * initial import * adding partial
deepset-ai · Jun 20, 2024 · 252e9d1 · 252e9d1
1 parent e208798
commit 252e9d1
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/haystack_experimental/evaluation/harness/rag/evaluation_pipeline.py b/haystack_experimental/evaluation/harness/rag/evaluation_pipeline.py
@@ -44,8 +44,8 @@ def default_rag_evaluation_pipeline(
         RAGEvaluationMetric.SEMANTIC_ANSWER_SIMILARITY: partial(
             SASEvaluator, model="sentence-transformers/all-MiniLM-L6-v2"
         ),
-        RAGEvaluationMetric.ANSWER_FAITHFULNESS: FaithfulnessEvaluator,
-        RAGEvaluationMetric.CONTEXT_RELEVANCE: ContextRelevanceEvaluator,
+        RAGEvaluationMetric.ANSWER_FAITHFULNESS: partial(FaithfulnessEvaluator, raise_on_failure=False),
+        RAGEvaluationMetric.CONTEXT_RELEVANCE: partial(ContextRelevanceEvaluator, raise_on_failure=False),
     }
 
     for metric in metrics: