diff --git a/ppocr/modeling/transforms/tps.py b/ppocr/modeling/transforms/tps.py
index 78338edf67d69e32322912d75dec01ce1e63cb49..dcce6246ac64b4b84229cbd69a4dc53c658b4c7b 100644
--- a/ppocr/modeling/transforms/tps.py
+++ b/ppocr/modeling/transforms/tps.py
@@ -230,15 +230,8 @@ class GridGenerator(nn.Layer):
     def build_inv_delta_C_paddle(self, C):
         """ Return inv_delta_C which is needed to calculate T """
         F = self.F
-        hat_C = paddle.zeros((F, F), dtype='float64')  # F x F
-        for i in range(0, F):
-            for j in range(i, F):
-                if i == j:
-                    hat_C[i, j] = 1
-                else:
-                    r = paddle.norm(C[i] - C[j])
-                    hat_C[i, j] = r
-                    hat_C[j, i] = r
+        hat_eye = paddle.eye(F, dtype='float64')  # F x F
+        hat_C = paddle.norm(C.reshape([1, F, 2]) - C.reshape([F, 1, 2]), axis=2) + hat_eye
         hat_C = (hat_C**2) * paddle.log(hat_C)
         delta_C = paddle.concat(  # F+3 x F+3
             [
diff --git a/tools/infer/predict_det.py b/tools/infer/predict_det.py
index 1b52e717ca1edbbd9ede31260e47ec8973270d3f..b284654c23b04f9c959ca8b55d8b91ccf0236e05 100755
--- a/tools/infer/predict_det.py
+++ b/tools/infer/predict_det.py
@@ -237,3 +237,4 @@ if __name__ == "__main__":
                                 "det_res_{}".format(img_name_pure))
         cv2.imwrite(img_path, src_im)
         logger.info("The visualized image saved in {}".format(img_path))
+
diff --git a/tools/infer/predict_rec.py b/tools/infer/predict_rec.py
index 0d847046530c02c9b0591bb4b379fd7ddeac1263..c5669e4bf1c381337af8f0decc6ff682bd225c6a 100755
--- a/tools/infer/predict_rec.py
+++ b/tools/infer/predict_rec.py
@@ -322,7 +322,8 @@ def main(args):
             'total_time_s': rec_time_dict['total_time']
         }
         benchmark_log = benchmark_utils.PaddleInferBenchmark(
-            text_recognizer.config, model_info, data_info, perf_info, mems)
+            text_recognizer.config, model_info, data_info, perf_info, mems,
+            args.save_log_path)
         benchmark_log("Rec")
 
 
diff --git a/tools/infer/utility.py b/tools/infer/utility.py
index 90ac5aa5ba2a33707965159d0486bb42957e3fce..7a70aec6f27a50aa8fa710d60bb1c23fe28e1bd3 100755
--- a/tools/infer/utility.py
+++ b/tools/infer/utility.py
@@ -37,6 +37,7 @@ def init_args():
     parser.add_argument("--use_gpu", type=str2bool, default=True)
     parser.add_argument("--ir_optim", type=str2bool, default=True)
     parser.add_argument("--use_tensorrt", type=str2bool, default=False)
+    parser.add_argument("--min_subgraph_size", type=int, default=3)
     parser.add_argument("--precision", type=str, default="fp32")
     parser.add_argument("--gpu_mem", type=int, default=500)
 
@@ -236,12 +237,14 @@ def create_predictor(args, mode, logger):
             config.enable_tensorrt_engine(
                 precision_mode=inference.PrecisionType.Float32,
                 max_batch_size=args.max_batch_size,
-                min_subgraph_size=3)  # skip the minmum trt subgraph
-        if mode == "det" and "mobile" in model_file_path:
+                min_subgraph_size=args.min_subgraph_size)
+            # skip the minmum trt subgraph
+        if mode == "det":
             min_input_shape = {
                 "x": [1, 3, 50, 50],
                 "conv2d_92.tmp_0": [1, 96, 20, 20],
                 "conv2d_91.tmp_0": [1, 96, 10, 10],
+                "conv2d_59.tmp_0": [1, 96, 20, 20],
                 "nearest_interp_v2_1.tmp_0": [1, 96, 10, 10],
                 "nearest_interp_v2_2.tmp_0": [1, 96, 20, 20],
                 "nearest_interp_v2_3.tmp_0": [1, 24, 20, 20],
@@ -254,6 +257,7 @@ def create_predictor(args, mode, logger):
                 "x": [1, 3, 2000, 2000],
                 "conv2d_92.tmp_0": [1, 96, 400, 400],
                 "conv2d_91.tmp_0": [1, 96, 200, 200],
+                "conv2d_59.tmp_0": [1, 96, 400, 400],
                 "nearest_interp_v2_1.tmp_0": [1, 96, 200, 200],
                 "nearest_interp_v2_2.tmp_0": [1, 96, 400, 400],
                 "nearest_interp_v2_3.tmp_0": [1, 24, 400, 400],
@@ -266,6 +270,7 @@ def create_predictor(args, mode, logger):
                 "x": [1, 3, 640, 640],
                 "conv2d_92.tmp_0": [1, 96, 160, 160],
                 "conv2d_91.tmp_0": [1, 96, 80, 80],
+                "conv2d_59.tmp_0": [1, 96, 160, 160],
                 "nearest_interp_v2_1.tmp_0": [1, 96, 80, 80],
                 "nearest_interp_v2_2.tmp_0": [1, 96, 160, 160],
                 "nearest_interp_v2_3.tmp_0": [1, 24, 160, 160],
@@ -274,31 +279,6 @@ def create_predictor(args, mode, logger):
                 "elementwise_add_7": [1, 56, 40, 40],
                 "nearest_interp_v2_0.tmp_0": [1, 96, 40, 40]
             }
-        if mode == "det" and "server" in model_file_path:
-            min_input_shape = {
-                "x": [1, 3, 50, 50],
-                "conv2d_59.tmp_0": [1, 96, 20, 20],
-                "nearest_interp_v2_2.tmp_0": [1, 96, 20, 20],
-                "nearest_interp_v2_3.tmp_0": [1, 24, 20, 20],
-                "nearest_interp_v2_4.tmp_0": [1, 24, 20, 20],
-                "nearest_interp_v2_5.tmp_0": [1, 24, 20, 20]
-            }
-            max_input_shape = {
-                "x": [1, 3, 2000, 2000],
-                "conv2d_59.tmp_0": [1, 96, 400, 400],
-                "nearest_interp_v2_2.tmp_0": [1, 96, 400, 400],
-                "nearest_interp_v2_3.tmp_0": [1, 24, 400, 400],
-                "nearest_interp_v2_4.tmp_0": [1, 24, 400, 400],
-                "nearest_interp_v2_5.tmp_0": [1, 24, 400, 400]
-            }
-            opt_input_shape = {
-                "x": [1, 3, 640, 640],
-                "conv2d_59.tmp_0": [1, 96, 160, 160],
-                "nearest_interp_v2_2.tmp_0": [1, 96, 160, 160],
-                "nearest_interp_v2_3.tmp_0": [1, 24, 160, 160],
-                "nearest_interp_v2_4.tmp_0": [1, 24, 160, 160],
-                "nearest_interp_v2_5.tmp_0": [1, 24, 160, 160]
-            }
         elif mode == "rec":
             min_input_shape = {"x": [args.rec_batch_num, 3, 32, 10]}
             max_input_shape = {"x": [args.rec_batch_num, 3, 32, 2000]}