fix: The randomness of the local vector model cannot be used (#4613)

Author: shaohuzhang1

apps/models_provider/impl/local_model_provider/model/embedding/model.py

@@ -10,8 +10,11 @@
 
 from langchain_huggingface import HuggingFaceEmbeddings
 
+from common.utils.logger import maxkb_logger
 from models_provider.base_model_provider import MaxKBBaseModel
 
+max_retries = 3
+
 
 class LocalEmbedding(MaxKBBaseModel, HuggingFaceEmbeddings):
     @staticmethod
@@ -20,7 +23,20 @@ def is_cache_model():
 
     @staticmethod
     def new_instance(model_type, model_name, model_credential: Dict[str, object], **model_kwargs):
-        return LocalEmbedding(model_name=model_name, cache_folder=model_credential.get('cache_folder'),
-                              model_kwargs={'device': model_credential.get('device')},
-                              encode_kwargs={'normalize_embeddings': True}
-                              )
+        for attempt in range(max_retries):
+            try:
+                embedding = LocalEmbedding(model_name=model_name, cache_folder=model_credential.get('cache_folder'),
+                                           model_kwargs={'device': model_credential.get('device')},
+                                           encode_kwargs={'normalize_embeddings': True}
+                                           )
+                # 测试一下是否真的能用
+                embedding.embed_query("test")
+                return embedding
+            except Exception as e:
+                if 'meta tensor' in str(e).lower() and attempt < max_retries - 1:
+                    maxkb_logger.warning(
+                        f"Test failed with meta tensor error, retrying... (attempt {attempt + 1}/{max_retries})")
+                    import time
+                    time.sleep(1)
+                    continue
+                raise e