# Triton backend to use name: "yolo-ref-branch" platform: "onnxruntime_onnx" max_batch_size: 0 # Triton should expect as input a single string # input of variable length named 'text_input' input [ { name: "images" data_type: TYPE_FP32 dims: [1, 3, 640, 640] } ] # Triton should expect to respond with a single string # output of variable length named 'text_output' output [ { name: "output0" data_type: TYPE_FP32 dims: [1, 84, 8400] } ] instance_group [ { kind: KIND_AUTO count: 1 } ]