diff --git a/scripts/vsmlrt.py b/scripts/vsmlrt.py
index debf834..1d2ea13 100644
--- a/scripts/vsmlrt.py
+++ b/scripts/vsmlrt.py
@@ -132,7 +132,7 @@ class TRT:
         output_format: int = 0 # 0: fp32, 1: fp16
         min_shapes: typing.Tuple[int, int] = (0, 0)
         faster_dynamic_shapes: bool = True
-        obey_fp16: bool = False
+        force_fp16: bool = False
 
         # internal backend attributes
         supports_onnx_serialization: bool = False
@@ -1039,7 +1039,7 @@ def trtexec(
     output_format: int = 0,
     min_shapes: typing.Tuple[int, int] = (0, 0),
     faster_dynamic_shapes: bool = True,
-    obey_fp16: bool = False
+    force_fp16: bool = False
 ) -> str:
 
     # tensort runtime version, e.g. 8401 => 8.4.1
@@ -1051,7 +1051,7 @@ def trtexec(
     if isinstance(max_shapes, int):
         max_shapes = (max_shapes, max_shapes)
 
-    if obey_fp16:
+    if force_fp16:
         fp16 = True
         tf32 = False
 
@@ -1158,7 +1158,7 @@ def trtexec(
     if faster_dynamic_shapes and not static_shape and trt_version >= 8500:
         args.append("--preview=+fasterDynamicShapes0805")
 
-    if obey_fp16:
+    if force_fp16:
         if trt_version >= 8401:
             args.extend([
                 "--layerPrecisions=*:fp16",
@@ -1166,7 +1166,7 @@ def trtexec(
                 "--precisionConstraints=obey"
             ])
         else:
-            raise ValueError('"obey_fp16" is not available')
+            raise ValueError('"force_fp16" is not available')
 
     if log:
         env_key = "TRTEXEC_LOG_FILE"
@@ -1376,7 +1376,7 @@ def _inference(
             output_format=backend.output_format,
             min_shapes=backend.min_shapes,
             faster_dynamic_shapes=backend.faster_dynamic_shapes,
-            obey_fp16=backend.obey_fp16
+            force_fp16=backend.force_fp16
         )
         clip = core.trt.Model(
             clips, engine_path,
@@ -1502,7 +1502,6 @@ class BackendV2:
     @staticmethod
     def TRT(*,
         num_streams: int = 1,
-        obey_fp16: bool = False,
         fp16: bool = False,
         tf32: bool = True,
         output_format: int = 0, # 0: fp32, 1: fp16
@@ -1512,6 +1511,7 @@ def TRT(*,
         min_shapes: typing.Tuple[int, int] = (0, 0),
         opt_shapes: typing.Optional[typing.Tuple[int, int]] = None,
         max_shapes: typing.Optional[typing.Tuple[int, int]] = None,
+        force_fp16: bool = False,
         use_cublas: bool = False,
         use_cudnn: bool = True,
         device_id: int = 0,
@@ -1520,7 +1520,7 @@ def TRT(*,
 
         return Backend.TRT(
             num_streams=num_streams,
-            fp16=fp16, obey_fp16=obey_fp16, tf32=tf32, output_format=output_format,
+            fp16=fp16, force_fp16=force_fp16, tf32=tf32, output_format=output_format,
             workspace=workspace, use_cuda_graph=use_cuda_graph,
             static_shape=static_shape,
             min_shapes=min_shapes, opt_shapes=opt_shapes, max_shapes=max_shapes,