fix Rerank Vector Similarity Score (infiniflow#1249)

### What problem does this PR solve?

infiniflow#1243 
### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

rag/llm/rerank_model.py

@@ -71,6 +71,7 @@ def similarity(self, query: str, texts: list):
         res = []
         for i in range(0, len(pairs), batch_size):
             scores = self._model.compute_score(pairs[i:i + batch_size], max_length=2048)
+            scores = sigmoid(np.array(scores)).tolist()
             if isinstance(scores, float): res.append(scores)
             else:  res.extend(scores)
         return np.array(res), token_count