Update app.py

app.py

@@ -24,7 +24,7 @@ model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloa
 @torch.inference_mode()
 @spaces.GPU
 def predict_math_bot(user_message, system_message="", max_new_tokens=125, temperature=0.1, top_p=0.9, repetition_penalty=1.9, do_sample=False):
-    prompt = f"<s><INST>{user_message}{system_message}<|\INST|>" if system_message else user_message
+    prompt = f"<s><INST>{user_message}{system_message}<\INST>" if system_message else user_message
     inputs = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
     input_ids = inputs["input_ids"].to(model.device)