Update app.py

app.py

@@ -25,7 +25,7 @@ model = AutoModelForCausalLM.from_pretrained(
   device_map="cuda", 
   trust_remote_code=True, 
   torch_dtype="auto", 
-  _attn_implementation='flash_attention_2',
+#  _attn_implementation='flash_attention_2',
   load_in_4bit=True # Optional: Load model in 4-bit mode to save memory
 )