add is_size_fixed member

Add is_size_fixed member to check whether output tensor size should be updated
or not. In case of Hailo NPU backend, there are some caces that output tensor
size cannot be calculated by tensor shape such as yolov10 models. The output
tensor size of These models is calculated hailo runtime internally.

Change-Id: I92e31b5ceea5a85f61c846578797deef0998d2e4
Signed-off-by: Inki Dae <inki.dae@samsung.com>

diff --git a/include/inference_engine_type.h b/include/inference_engine_type.h
index 11016a746a4afd9cfab70ada00107bf495d2dd17..bf05487be121979b5fdf1838d08baf4392997b0c 100644 (file)
--- a/include/inference_engine_type.h
+++ b/include/inference_engine_type.h
@@ -228,6 +228,7 @@ extern "C"
                size_t size {}; /** tensor element size. The size should be height * width * channel count */
                float scale {}; /**< a scale value of the layer. */
                int zero_point {}; /**< a zero point value of the layer. */
+               bool is_size_fixed {}; /** < if true then tensor size is calculated inference engine internally so do not use 'shape' member to calculate the tensor size. Use 'size' member instead.*/
                inference_tensor_quantization_type_e quantization_type { INFERENCE_TENSOR_QUANTIZATION_NONE }; /**< a quantization type of the layer. */
                // TODO.
        } inference_engine_tensor_info;