Update agent.py

agent.py

@@ -161,10 +161,10 @@ try:
     logger.info(f"Используемое устройство: {device}")
 
     # Инициализация Qwen2-7B
-    qwen_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-1.5B-Instruct")
+    qwen_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-7B-Instruct") # ("Qwen/Qwen2-1.5B-Instruct")
     qwen_model = AutoModelForCausalLM.from_pretrained(
-        # "Qwen/Qwen2-7B-Instruct",
-        "Qwen/Qwen2-1.5B-Instruct",
+        "Qwen/Qwen2-7B-Instruct",
+        # "Qwen/Qwen2-1.5B-Instruct",
         device_map="auto",
         load_in_4bit=True if device == "cuda" else False,  # Квантование для GPU
         torch_dtype=torch.float16 if device == "cuda" else torch.float32
@@ -880,8 +880,8 @@ def create_answer(state: AgentState) -> AgentState:
         logger.info(f"Используемое устройство: {device}")
         qwen_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-7B-Instruct")
         qwen_model = AutoModelForCausalLM.from_pretrained(
-            # "Qwen/Qwen2-7B-Instruct",
-            "Qwen/Qwen2-1.5B-Instruct",
+            "Qwen/Qwen2-7B-Instruct",
+            # "Qwen/Qwen2-1.5B-Instruct",
             device_map="auto",
             torch_dtype=torch.bfloat16,  # Используем bfloat16 для экономии памяти
             low_cpu_mem_usage=True  # Оптимизация для CPU