Update app.py

app.py

@@ -34,8 +34,8 @@ cache_dir = "/data"
 # )
 model = Llama4ForConditionalGeneration.from_pretrained(
     model_name,
-    # flex_attention is only needed for image
-    # attn_implementation="flex_attention", 
+    # default is eager attention
+    attn_implementation="flex_attention",
     # gguf_file=filename,
     cache_dir = cache_dir,
     torch_dtype=torch_dtype,