fix: answer_correctness embedding (#513)

src/ragas/metrics/_answer_correctness.py

@@ -104,6 +104,8 @@ def __post_init__(self: t.Self):
         if not all([w >= 0 for w in self.weights]):
             raise ValueError("Weights must be non-negative")
 
+    def init_model(self):
+        super().init_model()
         if self.answer_similarity is None and self.weights[1] != 0:
             self.answer_similarity = AnswerSimilarity(
                 llm=self.llm, batch_size=self.batch_size, embeddings=self.embeddings

tests/benchmarks/benchmark_eval.py

@@ -19,7 +19,7 @@
 # data
 ds = load_dataset("explodinggradients/amnesty_qa", "english")
 assert isinstance(ds, DatasetDict)
-eval_dataset = ds["train"]
+eval_dataset = ds["eval"]
 
 # metrics
 metrics = [