테스트

app.py

@@ -14,8 +14,10 @@ from config.util.org_filter import extract_keywords_simple
 from repository.usersRepository import findAll
 from config.ai.call_llm_model import get_qwen_model
 from config.db.chroma import collection
-
+import torch_directml
+# DirectML 디바이스 선언
 store_data = {}
+device = torch_directml.device()
 
 def search_employees(data: List[dict], query: str) -> List[dict]:
     """
@@ -79,7 +81,6 @@ def query_select_summarize_stream(results, query, ai, min_similarity: float = 0.
     Returns:
         Generator: 스트리밍 응답 제너레이터
     """
-
     if ai :
         if not results['documents'] or not results['documents'][0]:
             def generate_empty():
@@ -135,7 +136,7 @@ def query_select_summarize_stream(results, query, ai, min_similarity: float = 0.
         enable_thinking=False, # Qwen 모델 버전에 따라 지원 여부 확인 필요
         return_tensors="pt",
         return_dict=True
-    ).to(model.device)
+    ).to(device)
 
     # 스트리머 설정
     streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
@@ -205,7 +206,7 @@ def query_summarize_simple(query: str) :
         add_generation_prompt=True,
         enable_thinking=False # Qwen 모델 버전에 따라 지원 여부 확인 필요
     )
-    model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
+    model_inputs = tokenizer([text], return_tensors="pt").to(device)
 
     # conduct text completion
     generated_ids = model.generate(