diff --git a/xinference/model/rerank/mindie.py b/xinference/model/rerank/mindie.py
index b49ef024d2..2f7e6c2da4 100644
--- a/xinference/model/rerank/mindie.py
+++ b/xinference/model/rerank/mindie.py
@@ -67,29 +67,20 @@ def rerank(
         attention_mask = encoded_input.data["attention_mask"].numpy().astype(np.int64)
         n_tokens = attention_mask.sum().item()
 
-        similarity_scores = []
-        for i in range(len(sentence_combinations)):
-            output = session.infer(
-                feeds=[input_ids, attention_mask],
-                mode="dymshape",
-                custom_sizes=10000000,
-            )
-            scores = (
-                torch.from_numpy(output[0][:, 0])
-                .view(
-                    -1,
-                )
-                .float()
-            )
-            exec_time = session.summary().exec_time_list[-1]
-            logger.info(
-                "%s%s inference time: %.2f ms",
-                i + 1,
-                "tsnrhtdd"[(i + 1) % 5 * ((i + 1) % 100 ^ 15 > 4 > (i + 1) % 10) :: 4],
-                exec_time[1] - exec_time[0],
+        output = session.infer(
+            feeds=[input_ids, attention_mask],
+            mode="dymshape",
+            custom_sizes=10000000,
+        )
+        scores = (
+            torch.from_numpy(output[0][:, 0])
+            .view(
+                -1,
             )
-            logger.info("scores [positive, negative]: %s", scores)
-            similarity_scores.append(float(scores[0]))
+            .float()
+        )
+        logger.info("scores [positive, negative]: %s", scores)
+        similarity_scores = scores.tolist()
 
         sim_scores_argsort = list(reversed(np.argsort(similarity_scores)))
         if top_n is not None:
@@ -132,7 +123,7 @@ def rerank(
         }
 
         del similarity_scores
-        self._session.free_resource()
+        # self._session.free_resource()
 
         return Rerank(id=str(uuid.uuid1()), results=docs, meta=metadata)