Update app.py

app.py

@@ -50,6 +50,7 @@ model = AutoModelForCausalLM.from_pretrained(
     MODEL,
     torch_dtype=torch.float16,
     device_map="auto",
+    attn_implementation="flash_attention_2",
     quantization_config=quantization_config)
 
 # Ensure `pad_token_id` is set