prepare_backend

dragon.vm.tensorrt.onnx.prepare_backend(
  model,
  device='CUDA:0',
  **kwargs
)

Create a backend to execute repeatedly.

Parameters:
  • model (onnx.ModelProto) – The onnx model.
  • device (str, optional) – The executing device.
Returns:

tensorrt.onnx.BackendRep – The backend.