Run inference with an uploaded ONNX model.