import Tensorrt not in compile time in order to dynamically load tensorrt only if nvidia gpu is present

Azaion.Inference/onnx_engine.pyx

@@ -14,11 +14,11 @@ cdef class OnnxEngine(InferenceEngine):
         model_meta = self.session.get_modelmeta()
         print("Metadata:", model_meta.custom_metadata_map)
 
-    cdef tuple get_input_shape(self):
+    cpdef tuple get_input_shape(self):
         shape = self.input_shape
         return shape[2], shape[3]
 
-    cdef int get_batch_size(self):
+    cpdef int get_batch_size(self):
         return self.batch_size
 
     cpdef run(self, input_data):