Add enable_mkldnn to static/deploy/python/infer.py (#5059)

* Add enable_mkldnn to infer.py

static/deploy/python/infer.py

@@ -59,6 +59,7 @@ class Detector(object):
         device (str): Choose the device you want to run, it can be: CPU/GPU/XPU, default is CPU
         run_mode (str): mode of running(fluid/trt_fp32/trt_fp16)
         threshold (float): threshold to reserve the result for output.
+        enable_mkldnn (bool): whether use mkldnn with CPU.
     """
 
     def __init__(self,
@@ -67,7 +68,8 @@ class Detector(object):
                  device='CPU',
                  run_mode='fluid',
                  threshold=0.5,
-                 trt_calib_mode=False):
+                 trt_calib_mode=False,
+                 enable_mkldnn=False):
         self.config = config
         if self.config.use_python_inference:
             self.executor, self.program, self.fecth_targets = load_executor(
@@ -78,7 +80,8 @@ class Detector(object):
                 run_mode=run_mode,
                 min_subgraph_size=self.config.min_subgraph_size,
                 device=device,
-                trt_calib_mode=trt_calib_mode)
+                trt_calib_mode=trt_calib_mode,
+                enable_mkldnn=enable_mkldnn)
 
     def preprocess(self, im):
         preprocess_ops = []
@@ -225,14 +228,16 @@ class DetectorSOLOv2(Detector):
                  device='CPU',
                  run_mode='fluid',
                  threshold=0.5,
-                 trt_calib_mode=False):
+                 trt_calib_mode=False,
+                 enable_mkldnn=False):
         super(DetectorSOLOv2, self).__init__(
             config=config,
             model_dir=model_dir,
             device=device,
             run_mode=run_mode,
             threshold=threshold,
-            trt_calib_mode=trt_calib_mode)
+            trt_calib_mode=trt_calib_mode,
+            enable_mkldn=enable_mkldnn)
 
     def predict(self,
                 image,
@@ -385,13 +390,15 @@ def load_predictor(model_dir,
                    batch_size=1,
                    device='CPU',
                    min_subgraph_size=3,
-                   trt_calib_mode=False):
+                   trt_calib_mode=False,
+                   enable_mkldnn=False):
     """set AnalysisConfig, generate AnalysisPredictor
     Args:
         model_dir (str): root path of __model__ and __params__
         device (str): Choose the device you want to run, it can be: CPU/GPU/XPU, default is CPU
         trt_calib_mode (bool): If the model is produced by TRT offline quantitative
             calibration, trt_calib_mode need to set True
+        enable_mkldnn (bool): Whether use mkldnn with CPU, default is False
     Returns:
         predictor (PaddlePredictor): AnalysisPredictor
     Raises:
@@ -419,7 +426,10 @@ def load_predictor(model_dir,
         config.enable_xpu(10 * 1024 * 1024)
     else:
         config.disable_gpu()
-
+        if enable_mkldnn:
+            config.set_mkldnn_cache_capacity(0)
+            config.enable_mkldnn()
+            config.pass_builder().append_pass("interpolate_mkldnn_pass")
     if run_mode in precision_map.keys():
         config.enable_tensorrt_engine(
             workspace_size=1 << 10,
@@ -432,6 +442,7 @@ def load_predictor(model_dir,
     # disable print log when predict
     config.disable_glog_info()
     # enable shared memory
+    if (not enable_mkldnn):
         config.enable_memory_optim()
     # disable feed, fetch OP, needed by zero_copy_run
     config.switch_use_feed_fetch_ops(False)
@@ -545,14 +556,16 @@ def main():
         FLAGS.model_dir,
         device=FLAGS.device,
         run_mode=FLAGS.run_mode,
-        trt_calib_mode=FLAGS.trt_calib_mode)
+        trt_calib_mode=FLAGS.trt_calib_mode,
+        enable_mkldnn=FLAGS.enable_mkldnn)
     if config.arch == 'SOLOv2':
         detector = DetectorSOLOv2(
             config,
             FLAGS.model_dir,
             device=FLAGS.device,
             run_mode=FLAGS.run_mode,
-            trt_calib_mode=FLAGS.trt_calib_mode)
+            trt_calib_mode=FLAGS.trt_calib_mode,
+            enable_mkldnn=FLAGS.enable_mkldnn)
     # predict from image
     if FLAGS.image_file != '':
         predict_image(detector)
@@ -618,7 +631,11 @@ if __name__ == '__main__':
         default=False,
         help="If the model is produced by TRT offline quantitative "
         "calibration, trt_calib_mode need to set True.")
-
+    parser.add_argument(
+        "--enable_mkldnn",
+        type=ast.literal_eval,
+        default=False,
+        help="Whether use mkldnn with CPU.")
     FLAGS = parser.parse_args()
     print_arguments(FLAGS)
     if FLAGS.image_file != '' and FLAGS.video_file != '':