read cdn yaml config from api

automate tensorrt model conversion in case of no existing one for user's gpu

Azaion.Inference/inference_engine.pxd

@@ -26,4 +26,13 @@ cdef class TensorRTEngine(InferenceEngine):
     cdef object input_shape
     cdef object output_shape
     cdef object h_output
-    cdef object class_names
+    cdef object class_names
+
+    @staticmethod
+    cdef bytes convert_from_onnx(bytes onnx_model)
+
+    @staticmethod
+    cdef unsigned long long get_gpu_memory_bytes(device_id=?)
+
+    @staticmethod
+    cdef str get_engine_filename(device_id=?)