split to 2 files: tensorrt_engine and onnx engine

Azaion.Inference/inference.pyx

@@ -2,6 +2,7 @@ import json
 import mimetypes
 import os
 import subprocess
+import sys
 import time
 
 import cv2
@@ -11,10 +12,15 @@ cimport constants
 from remote_command cimport RemoteCommand
 from annotation cimport Detection, Annotation
 from ai_config cimport AIRecognitionConfig
-from inference_engine cimport OnnxEngine, TensorRTEngine
 from hardware_service cimport HardwareService
 from security cimport Security
 
+if HardwareService.has_nvidia_gpu():
+    from tensorrt_engine cimport TensorRTEngine
+else:
+    from onnx_engine import OnnxEngine
+
+
 cdef class Inference:
     def __init__(self, api_client, on_annotation):
         self.api_client = api_client
@@ -31,17 +37,20 @@ cdef class Inference:
         if not is_nvidia:
             return
 
-        engine_filename = TensorRTEngine.get_engine_filename()
+        engine_filename = TensorRTEngine.get_engine_filename(0)
         key = Security.get_model_encryption_key()
         models_dir = constants.MODELS_FOLDER
         if not os.path.exists(os.path.join(<str> models_dir, f'{engine_filename}.big')):
+            #TODO: Check cdn on engine exists, if there is, download
             self.is_building_engine = True
-            time.sleep(5) # prevent simultaneously loading dll and models
+            time.sleep(8) # prevent simultaneously loading dll and models
             onnx_model = self.api_client.load_big_small_resource(constants.AI_ONNX_MODEL_FILE, models_dir, key)
             model_bytes = TensorRTEngine.convert_from_onnx(onnx_model)
             self.api_client.upload_big_small_resource(model_bytes, <str> engine_filename, models_dir, key)
             print('uploaded ')
             self.is_building_engine = False
+        else:
+            print('tensor rt engine is here, no need to build')
 
 
     cdef init_ai(self):
@@ -54,10 +63,9 @@ cdef class Inference:
         if is_nvidia:
             while self.is_building_engine:
                 time.sleep(1)
-            engine_filename = TensorRTEngine.get_engine_filename()
+            engine_filename = TensorRTEngine.get_engine_filename(0)
             model_bytes = self.api_client.load_big_small_resource(engine_filename, models_dir, key)
             self.engine = TensorRTEngine(model_bytes)
-
         else:
             model_bytes = self.api_client.load_big_small_resource(constants.AI_ONNX_MODEL_FILE, models_dir, key)
             self.engine = OnnxEngine(model_bytes)