Update app.py

app.py

@@ -32,7 +32,7 @@ def load_model():
         model = AutoModelForCausalLM.from_pretrained(
             model_name,
             device_map="auto",
-            torch_dtype="auto"  # A100 ortamında bf16 otomatik seçer
+            torch_dtype="auto"
             # Eğer istersen load_in_8bit=True parametresini ekleyebiliriz
         )
 
@@ -46,6 +46,7 @@ def load_model():
 
 class UserInputRequest(BaseModel):
     user_input: str
+    system_prompt: str
 
 @app.post("/generate")
 def generate(req: UserInputRequest):
@@ -53,13 +54,16 @@ def generate(req: UserInputRequest):
         start_time = time.time()
         log(f"💬 Kullanıcı isteği alındı: {req.user_input}")
 
+        # System prompt + user input birleştirme
+        full_prompt = f"{req.system_prompt}\n\n{req.user_input}"
+
         result = pipe(
-            req.user_input,
-            max_new_tokens=50,          # Azaltıldı
-            temperature=0.0,            # Deterministic
-            top_p=1.0,                  # Serbest sampling
-            repetition_penalty=1.0,     # Doğal tekrarlar izinli
-            do_sample=False             # Sampling kapalı, deterministik
+            full_prompt,
+            max_new_tokens=200,
+            temperature=0.0,
+            top_p=1.0,
+            repetition_penalty=1.0,
+            do_sample=False
         )
         answer = result[0]["generated_text"]