def handle()

in src/sagemaker_huggingface_inference_toolkit/handler_service.py [0:0]

• 25 lines of code
• 8 McCabe index (conditional complexity)

    def handle(self, data, context):
        """Handles an inference request with input data and makes a prediction.

        Args:
            data (obj): the request data.
            context (obj): metadata on the incoming request data.

        Returns:
            list[obj]: The return value from the Transformer.transform method,
                which is a serialized prediction result wrapped in a list if
                inference is successful. Otherwise returns an error message
                with the context set appropriately.

        """
        try:
            if not self.initialized:
                if self.attempted_init:
                    logger.warn(
                        "Model is not initialized, will try to load model again.\n"
                        "Please consider increase wait time for model loading.\n"
                    )
                self.initialize(context)

            input_data = data[0].get("body")

            request_property = context.request_processor[0].get_request_properties()
            content_type = utils.retrieve_content_type_header(request_property)
            accept = request_property.get("Accept") or request_property.get("accept")

            if not accept or accept == content_types.ANY:
                accept = content_types.JSON

            if content_type in content_types.UTF8_TYPES:
                input_data = input_data.decode("utf-8")

            predict_start = time.time()
            response = self.transform_fn(*([self.model, input_data, content_type, accept] + self.transform_extra_arg))
            predict_end = time.time()

            context.metrics.add_time("Transform Fn", round((predict_end - predict_start) * 1000, 2))

            context.set_response_content_type(0, accept)
            return [response]

        except Exception as e:
            raise PredictionException(str(e), 400)