def evaluate_inference_speed()

in source/neo/eval.py [0:0]

• 21 lines of code
• 6 McCabe index (conditional complexity)

    def evaluate_inference_speed(self, repeat_times, neo_optimized=True):
        """
        object detector's inference speed evaluation

        :param repeat_times: repeating times for test
        :param neo_optimized: whether to using sagemaker neo optimized model, boolean variable
        :return: None
        """
        image_data_for_raw_model, image_data_for_neo_optimized_model = self.get_model_input()
        _, channels, height, width = image_data_for_raw_model.shape

        if not neo_optimized:
            # warmup
            for i in range(100):
                _ = self.eval_without_neo(x=image_data_for_raw_model)

            # inference test
            t_start = time.time()
            for i in range(repeat_times):
                _ = self.eval_without_neo(x=image_data_for_raw_model)
            t_end = time.time()

            print('[NEO Optimization Disabled] Time Cost per Frame (input size = 1x{}x{}x{}) = {} ms'.format(
                channels, height, width, 1000 * (t_end - t_start) / repeat_times))
        else:
            # warmup
            for i in range(100):
                _ = self.eval_with_neo(x=image_data_for_neo_optimized_model)

            t_start = time.time()
            for i in range(repeat_times):
                _ = self.eval_with_neo(x=image_data_for_neo_optimized_model)
            t_end = time.time()

            print('[NEO Optimization Enabled] Time Cost per Frame (input size = 1x{}x{}x{}) = {} ms'.format(
                channels, height, width, 1000 * (t_end - t_start) / repeat_times))