There is darknet format, TensorRT format, ONNX format, Pytorch format. Also YOLOv5 doesn’t have a configuration file like the other older formats
Deepstream uses TensorRT as the backend inference engine.
TensorRT supports Caffe, onnx, and uff format.
You can also convert YOLOv5 with the TensorRT native API directly.
Some similar example can be found below: