diff --git a/comfy/model_management.py b/comfy/model_management.py
index b1afeb715..241706925 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -281,19 +281,23 @@ def load_model_gpu(model):
             vram_set_state = VRAMState.LOW_VRAM
 
     real_model = model.model
+    patch_model_to = None
     if vram_set_state == VRAMState.DISABLED:
         pass
     elif vram_set_state == VRAMState.NORMAL_VRAM or vram_set_state == VRAMState.HIGH_VRAM or vram_set_state == VRAMState.SHARED:
         model_accelerated = False
-        real_model.to(torch_dev)
+        patch_model_to = torch_dev
 
     try:
-        real_model = model.patch_model()
+        real_model = model.patch_model(device_to=patch_model_to)
     except Exception as e:
         model.unpatch_model()
         unload_model()
         raise e
 
+    if patch_model_to is not None:
+        real_model.to(torch_dev)
+
     if vram_set_state == VRAMState.NO_VRAM:
         device_map = accelerate.infer_auto_device_map(real_model, max_memory={0: "256MiB", "cpu": "16GiB"})
         accelerate.dispatch_model(real_model, device_map=device_map, main_device=torch_dev)
diff --git a/comfy/samplers.py b/comfy/samplers.py
index 50fda016d..9eee25a92 100644
--- a/comfy/samplers.py
+++ b/comfy/samplers.py
@@ -248,7 +248,10 @@ def sampling_function(model_function, x, timestep, uncond, cond, cond_scale, con
 
                 c['transformer_options'] = transformer_options
 
-                output = model_function(input_x, timestep_, **c).chunk(batch_chunks)
+                if 'model_function_wrapper' in model_options:
+                    output = model_options['model_function_wrapper'](model_function, {"input": input_x, "timestep": timestep_, "c": c, "cond_or_uncond": cond_or_uncond}).chunk(batch_chunks)
+                else:
+                    output = model_function(input_x, timestep_, **c).chunk(batch_chunks)
                 del input_x
 
                 model_management.throw_exception_if_processing_interrupted()
diff --git a/comfy/sd.py b/comfy/sd.py
index 7f7c06bc5..1f364dd1f 100644
--- a/comfy/sd.py
+++ b/comfy/sd.py
@@ -338,7 +338,7 @@ class ModelPatcher:
                     sd.pop(k)
         return sd
 
-    def patch_model(self):
+    def patch_model(self, device_to=None):
         model_sd = self.model_state_dict()
         for key in self.patches:
             if key not in model_sd:
@@ -350,7 +350,10 @@ class ModelPatcher:
             if key not in self.backup:
                 self.backup[key] = weight.to(self.offload_device)
 
-            temp_weight = weight.to(torch.float32, copy=True)
+            if device_to is not None:
+                temp_weight = weight.float().to(device_to, copy=True)
+            else:
+                temp_weight = weight.to(torch.float32, copy=True)
             out_weight = self.calculate_weight(self.patches[key], temp_weight, key).to(weight.dtype)
             set_attr(self.model, key, out_weight)
             del temp_weight
diff --git a/cuda_malloc.py b/cuda_malloc.py
index 382432215..faee91a34 100644
--- a/cuda_malloc.py
+++ b/cuda_malloc.py
@@ -37,7 +37,7 @@ def get_gpu_names():
         return set()
 
 def cuda_malloc_supported():
-    blacklist = {"GeForce GTX 960M", "GeForce GTX 950M", "GeForce 945M", "GeForce 940M", "GeForce 930M", "GeForce 920M", "GeForce 910M", "GeForce GTX 750", "GeForce GTX 745"}
+    blacklist = {"GeForce GTX 960", "GeForce GTX 950", "GeForce 945M", "GeForce 940M", "GeForce 930M", "GeForce 920M", "GeForce 910M", "GeForce GTX 750", "GeForce GTX 745"}
     try:
         names = get_gpu_names()
     except: