From 5fa6f2f151c4ff39aa9e9175e97f0778c1fbc2be Mon Sep 17 00:00:00 2001
From: Stephen Hu <stephenhu@seismic.com>
Date: Tue, 15 Jul 2025 14:04:58 +0800
Subject: [PATCH] Update embedding_model.py (#8836)

### What problem does this PR solve?

Remove useless covert for bge encode_queries

### Type of change

- [x] Performance Improvement
---
 rag/llm/embedding_model.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/rag/llm/embedding_model.py b/rag/llm/embedding_model.py
index 16a415bf7..98f9bcbc1 100644
--- a/rag/llm/embedding_model.py
+++ b/rag/llm/embedding_model.py
@@ -114,7 +114,7 @@ class DefaultEmbedding(Base):
 
     def encode_queries(self, text: str):
         token_count = num_tokens_from_string(text)
-        return self._model.encode_queries([text]).tolist()[0], token_count
+        return self._model.encode_queries([text], convert_to_numpy=False)[0][0].cpu().numpy(), token_count
 
 
 class OpenAIEmbed(Base):