Update app.py

app.py

@@ -28,16 +28,16 @@ def predict(message, history, max_tokens, temperature, top_p):
     # Генерация ответа
     outputs = model.generate(
         input_ids=inputs, 
-        max_new_tokens=256,
+        max_new_tokens=max_tokens,
         do_sample=True,
-        temperature=0.7,
-        top_p=0.95,
+        temperature=temperature,
+        top_p=top_p,
     )
     
     # Декодирование результата
-    response = tokenizer.decode(outputs[0][len(inputs[0] + 9):], skip_special_tokens=True)
+    response = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
     
-    return response.strip()
+    return response.strip().replace("assistant", "", 1)
 
 # Настройка интерфейса Gradio
 iface = gr.ChatInterface(