diff --git a/ip_adapter/ip_adapter.py b/ip_adapter/ip_adapter.py
index ef6d05c..ec70eb5 100644
--- a/ip_adapter/ip_adapter.py
+++ b/ip_adapter/ip_adapter.py
@@ -204,7 +204,7 @@ class IPAdapter:
             prompt_embeds = torch.cat([prompt_embeds_, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds_, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
@@ -402,7 +402,7 @@ class IPAdapterPlusXL(IPAdapter):
             prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
diff --git a/ip_adapter/ip_adapter_faceid.py b/ip_adapter/ip_adapter_faceid.py
index 27e3b64..fe98ad5 100644
--- a/ip_adapter/ip_adapter_faceid.py
+++ b/ip_adapter/ip_adapter_faceid.py
@@ -238,7 +238,7 @@ class IPAdapterFaceID:
             prompt_embeds = torch.cat([prompt_embeds_, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds_, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
@@ -391,7 +391,7 @@ class IPAdapterFaceIDPlus:
             prompt_embeds = torch.cat([prompt_embeds_, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds_, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
@@ -456,7 +456,7 @@ class IPAdapterFaceIDXL(IPAdapterFaceID):
             prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
@@ -526,7 +526,7 @@ class IPAdapterFaceIDPlusXL(IPAdapterFaceIDPlus):
             prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
diff --git a/ip_adapter/ip_adapter_faceid_separate.py b/ip_adapter/ip_adapter_faceid_separate.py
index 8881d8b..80ca84c 100644
--- a/ip_adapter/ip_adapter_faceid_separate.py
+++ b/ip_adapter/ip_adapter_faceid_separate.py
@@ -246,7 +246,7 @@ class IPAdapterFaceID:
             prompt_embeds = torch.cat([prompt_embeds_, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds_, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
@@ -396,7 +396,7 @@ class IPAdapterFaceIDPlus:
             prompt_embeds = torch.cat([prompt_embeds_, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds_, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
@@ -461,7 +461,7 @@ class IPAdapterFaceIDXL(IPAdapterFaceID):
             prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
@@ -531,7 +531,7 @@ class IPAdapterFaceIDPlusXL(IPAdapterFaceIDPlus):
             prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
             negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
 
-        generator = get_generator(seed)
+        generator = get_generator(seed, self.device)
 
         images = self.pipe(
             prompt_embeds=prompt_embeds,
diff --git a/ip_adapter/utils.py b/ip_adapter/utils.py
index 46c446c..6a27335 100644
--- a/ip_adapter/utils.py
+++ b/ip_adapter/utils.py
@@ -80,13 +80,13 @@ def attnmaps2images(net_attn_maps):
 def is_torch2_available():
     return hasattr(F, "scaled_dot_product_attention")
 
-def get_generator(seed):
+def get_generator(seed, device):
 
     if seed is not None:
         if isinstance(seed, list):
-            generator = [torch.Generator(self.device).manual_seed(seed_item) for seed_item in seed]
+            generator = [torch.Generator(device).manual_seed(seed_item) for seed_item in seed]
         else:
-            generator = torch.Generator(self.device).manual_seed(seed)
+            generator = torch.Generator(device).manual_seed(seed)
     else:
         generator = None