split to 2 files: tensorrt_engine and onnx engine

Azaion.Inference/tensorrt_engine.pxd

@@ -0,0 +1,32 @@
+from inference_engine cimport InferenceEngine
+
+
+cdef class TensorRTEngine(InferenceEngine):
+
+    cdef public object context
+
+    cdef public object d_input
+    cdef public object d_output
+    cdef str input_name
+    cdef object input_shape
+
+    cdef object h_output
+    cdef str output_name
+    cdef object output_shape
+
+    cdef object stream
+
+    @staticmethod
+    cdef get_gpu_memory_bytes(int device_id)
+
+    @staticmethod
+    cdef get_engine_filename(int device_id)
+
+    @staticmethod
+    cdef convert_from_onnx(bytes onnx_model)
+
+    cdef tuple get_input_shape(self)
+
+    cdef int get_batch_size(self)
+
+    cdef run(self, input_data)