diff --git a/comfy/model_management.py b/comfy/model_management.py
index 803aa8a2a..810cfbff8 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import logging
 import sys
+import warnings
 from enum import Enum
 from threading import RLock
 from typing import Literal
@@ -101,7 +102,11 @@ def get_torch_device():
         if is_intel_xpu():
             return torch.device("xpu", torch.xpu.current_device())
         else:
-            return torch.device(torch.cuda.current_device())
+            try:
+                return torch.device(torch.cuda.current_device())
+            except:
+                warnings.warn("torch.cuda.current_device() did not return a device, returning a CPU torch device")
+                return torch.device("cpu")
 
 
 def get_total_memory(dev=None, torch_total_too=False):