Update llm_model.py

llm_model.py

@@ -3,14 +3,15 @@ from transformers import AutoTokenizer, AutoModelForCausalLM, AutoModelForSequen
 from log import log
 from pydantic import BaseModel
 
+model = None
+tokenizer = None
+eos_token_id = None
+
 class Message(BaseModel):
     user_input: str
 
 def setup_model(s_config):
     global model, tokenizer, eos_token_id
-    model = None
-    tokenizer = None
-    eos_token_id = None
     try:
         log("🧠 setup_model() başladı")
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")