Method load

quickstart/IntroNotebooks/onnx_helper.py:36–41

(self, file)

        self.stream = None
      
    def load(self, file):
        f = open(file, "rb")
        runtime = trt.Runtime(trt.Logger(trt.Logger.WARNING)) 

        engine = runtime.deserialize_cuda_engine(f.read())
        self.context = engine.create_execution_context()
        
    def allocate_memory(self, batch):
        self.output = np.empty(self.num_classes, dtype = self.target_dtype) # Need to set both input and output precisions to FP16 to fully enable FP16