def saveONNX(model, filepath):
  '''
  保存ONNX模型
  :param model: 神經(jīng)網(wǎng)絡(luò)模型
  :param filepath: 文件保存路徑
  '''
  
  # 神經(jīng)網(wǎng)絡(luò)輸入數(shù)據(jù)類型
  dummy_input = torch.randn(self.config.BATCH_SIZE, 1, 28, 28, device='cuda')
  torch.onnx.export(model, dummy_input, filepath, verbose=True)

 def ONNX_build_engine(onnx_file_path):
  '''
  通過加載onnx文件，構(gòu)建engine
  :param onnx_file_path: onnx文件路徑
  :return: engine
  '''
  # 打印日志
  G_LOGGER = trt.Logger(trt.Logger.WARNING)

  with trt.Builder(G_LOGGER) as builder, builder.create_network() as network, trt.OnnxParser(network, G_LOGGER) as parser:
   builder.max_batch_size = 100
   builder.max_workspace_size = 1 << 20

   print('Loading ONNX file from path {}...'.format(onnx_file_path))
   with open(onnx_file_path, 'rb') as model:
    print('Beginning ONNX file parsing')
    parser.parse(model.read())
   print('Completed parsing of ONNX file')

   print('Building an engine from file {}; this may take a while...'.format(onnx_file_path))
   engine = builder.build_cuda_engine(network)
   print("Completed creating Engine")

   # 保存計(jì)劃文件
   # with open(engine_file_path, "wb") as f:
   #  f.write(engine.serialize())
   return engine

 def loadONNX2TensorRT(filepath):
  '''
  通過onnx文件，構(gòu)建TensorRT運(yùn)行引擎
  :param filepath: onnx文件路徑
  '''
  # 計(jì)算開始時(shí)間
  Start = time()

  engine = self.ONNX_build_engine(filepath)

  # 讀取測試集
  datas = DataLoaders()
  test_loader = datas.testDataLoader()
  img, target = next(iter(test_loader))
  img = img.numpy()
  target = target.numpy()

  img = img.ravel()

  context = engine.create_execution_context()
  output = np.empty((100, 10), dtype=np.float32)

  # 分配內(nèi)存
  d_input = cuda.mem_alloc(1 * img.size * img.dtype.itemsize)
  d_output = cuda.mem_alloc(1 * output.size * output.dtype.itemsize)
  bindings = [int(d_input), int(d_output)]

  # pycuda操作緩沖區(qū)
  stream = cuda.Stream()
  # 將輸入數(shù)據(jù)放入device
  cuda.memcpy_htod_async(d_input, img, stream)
  # 執(zhí)行模型
  context.execute_async(100, bindings, stream.handle, None)
  # 將預(yù)測結(jié)果從從緩沖區(qū)取出
  cuda.memcpy_dtoh_async(output, d_output, stream)
  # 線程同步
  stream.synchronize()

  print("Test Case: " + str(target))
  print("Prediction: " + str(np.argmax(output, axis=1)))
  print("tensorrt time:", time() - Start)

  del context
  del engine