diff --git a/scripts/vsmlrt.py b/scripts/vsmlrt.py
index c5f3e52..ef47448 100644
--- a/scripts/vsmlrt.py
+++ b/scripts/vsmlrt.py
@@ -1,4 +1,4 @@
-__version__ = "3.20.6"
+__version__ = "3.20.7"
 
 __all__ = [
     "Backend", "BackendV2",
@@ -87,6 +87,7 @@ class ORT_CUDA:
         fp16: bool = False
         use_cuda_graph: bool = False # preview, not supported by all models
         fp16_blacklist_ops: typing.Optional[typing.Sequence[str]] = None
+        prefer_nhwc: bool = False
 
         # internal backend attributes
         supports_onnx_serialization: bool = True
@@ -2032,6 +2033,17 @@ def _inference(
             fp16_blacklist_ops=backend.fp16_blacklist_ops
         )
     elif isinstance(backend, Backend.ORT_CUDA):
+        kwargs = dict()
+
+        version_list = core.ort.Version().get("onnxruntime_version", b"0.0.0").split(b'.')
+        if len(version_list) != 3:
+            version = (0, 0, 0)
+        else:
+            version = tuple(map(int, version_list))
+
+        if version >= (1, 18, 0):
+            kwargs["prefer_nhwc"] = backend.prefer_nhwc
+
         clip = core.ort.Model(
             clips, network_path,
             overlap=overlap, tilesize=tilesize,
@@ -2043,7 +2055,8 @@ def _inference(
             fp16=backend.fp16,
             path_is_serialization=path_is_serialization,
             use_cuda_graph=backend.use_cuda_graph,
-            fp16_blacklist_ops=backend.fp16_blacklist_ops
+            fp16_blacklist_ops=backend.fp16_blacklist_ops,
+            **kwargs
         )
     elif isinstance(backend, Backend.OV_CPU):
         version = tuple(map(int, core.ov.Version().get("openvino_version", b"0.0.0").split(b'-')[0].split(b'.')))