Update main.py

main.py

@@ -17,7 +17,7 @@ model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     checkpoint,
     torch_dtype=torch.bfloat16,
     device_map="cuda",
-    attn_implementation="flash_attention_2",
+    # attn_implementation="flash_attention_2",
 )
 
 @app.get("/")