name: "parse_gui" platform: "ensemble" input [ { name: "IMAGE_BYTES" data_type: TYPE_STRING dims: [1] } ] output [ { name: "bbox" data_type: TYPE_FP32 dims: [-1, 4] }, { name: "text_embedding" data_type: TYPE_FP32 dims: [-1, 1024] }, { name: "function_embedding" data_type: TYPE_FP32 dims: [-1, 1024] }, { name: "vision_embedding" data_type: TYPE_FP32 dims: [-1, 768] } ] ensemble_scheduling { step { model_name: "parse_gui_bls" model_version: -1 input_map { key: "IMAGE_BYTES" value: "IMAGE_BYTES" } output_map { key: "bbox" value: "bbox" } output_map { key: "text_embedding" value: "text_embedding" } output_map { key: "function_embedding" value: "function_embedding" } output_map { key: "vision_embedding" value: "vision_embedding" } } }