Command option to set different devices for extensions

comfy/cli_args.py

@@ -43,6 +43,7 @@ parser.add_argument("--input-directory", type=str, default=None, help="Set the C
 parser.add_argument("--auto-launch", action="store_true", help="Automatically launch ComfyUI in the default browser.")
 parser.add_argument("--disable-auto-launch", action="store_true", help="Disable auto launching the browser.")
 parser.add_argument("--cuda-device", type=int, default=None, metavar="DEVICE_ID", help="Set the id of the cuda device this instance will use.")
+parser.add_argument("--extension-device", type=str, default=None, help="Set the device for extensions in the format 'extension:device;extension:device;...'.")
 cm_group = parser.add_mutually_exclusive_group()
 cm_group.add_argument("--cuda-malloc", action="store_true", help="Enable cudaMallocAsync (enabled by default for torch 2.0 and up).")
 cm_group.add_argument("--disable-cuda-malloc", action="store_true", help="Disable cudaMallocAsync.")

comfy/model_management.py

@@ -41,6 +41,12 @@ if args.directml is not None:
     # torch_directml.disable_tiled_resources(True)
     lowvram_available = False #TODO: need to find a way to get free memory in directml before this can be enabled by default.
 
+extensions_devices = {}
+if args.extension_device is not None:
+    for ext_dev in args.extension_device.split(";"):
+        ext, dev = ext_dev.split(":")
+        extensions_devices[ext] = dev
+
 try:
     import intel_extension_for_pytorch as ipex
     if torch.xpu.is_available():
@@ -69,6 +75,12 @@ def is_intel_xpu():
 def get_torch_device():
     global directml_enabled
     global cpu_state
+    global extensions_devices
+
+    extension = comfy.utils.get_extension_calling()
+    if extension is not None and extension in extensions_devices:
+        return torch.device(extensions_devices[extension])
+
     if directml_enabled:
         global directml_device
         return directml_device

comfy/utils.py

@@ -4,8 +4,19 @@ import struct
 import comfy.checkpoint_pickle
 import safetensors.torch
 import numpy as np
+import inspect
+import re
 from PIL import Image
 
+def get_extension_calling():
+    for frame in inspect.stack():
+        if "/custom_nodes/" in frame.filename:
+            stack_module = inspect.getmodule(frame[0]) 
+            if stack_module:
+                return re.sub(r".*\.?custom_nodes\.([^\.]+).*", r"\1", stack_module.__name__).split(".")[0]
+
+    return None
+
 def load_torch_file(ckpt, safe_load=False, device=None):
     if device is None:
         device = torch.device("cpu")