Are there any plans to adapt node next version for torch 2.10+?

[okolenmi]

There are a lot problems because more and more nodes want torch >2.6.0. Also the latest available torch XPU version is far ahead (>2.12). So it would be a logical step to work on this migration as I myself faced a lot of issues. My Ubuntu 26 (dev) doesn't support torch 2.6 by default. There are no compatible python version to download in repositories, so an average user should to do a lot of things to install a single node...

[openvino-dev-samples]

Hi @okolenmi Yes you are right. I plan to do a few investigation on torch 2.10 support. Will keep you tuned.

[okolenmi]

That's great news! When this extension works I have +70% to generation speed. It's slightly uncomfortable to wait for pre-processing (especially on high resolution), but still it have a lot of use cases for me and, I'm sure, for a lot of other people. It's just problematic to use it in its current state on modern dependencies. Thank you for your work.

[openvino-dev-samples]

I have updated the node to support torch 2.10. Maybe you can help to verify, thanks a lot!

[RoyalXXX]

I have updated the node to support torch 2.10. Maybe you can help to verify, thanks a lot!

Unfortunately, after updating ComfyUI-OpenVINO, it stopped working for me.

I attached what’s being output in my console. error.txt

According to Grok:

This error (AssertionError: fake mode ... from tracing context 0 doesn't match mode ... from active fake mode) is a classic PyTorch Dynamo + FakeTensorMode conflict. It happens when two different FakeTensorMode objects are active in the same tracing session.
In your case, the culprit is ComfyUI-OpenVINO (specifically its torch dynamo backend / openvino compiler backend). The OpenVINO backend explicitly creates its own FakeTensorMode here:

with FakeTensorMode(allow_non_fake_inputs=True) as fakemode:
    model = make_fx(subgraph, ...)(*fake_inputs)

Meanwhile, PyTorch's dynamo / torch.compile (used inside ComfyUI's sampler / model patching) already has its own active FakeTensorMode higher up the call stack. → Two modes clash → assertion crash.

[openvino-dev-samples]

I have updated the node to support torch 2.10. Maybe you can help to verify, thanks a lot!

Unfortunately, after updating ComfyUI-OpenVINO, it stopped working for me.

I attached what’s being output in my console. error.txt

According to Grok:

This error (AssertionError: fake mode ... from tracing context 0 doesn't match mode ... from active fake mode) is a classic PyTorch Dynamo + FakeTensorMode conflict. It happens when two different FakeTensorMode objects are active in the same tracing session. In your case, the culprit is ComfyUI-OpenVINO (specifically its torch dynamo backend / openvino compiler backend). The OpenVINO backend explicitly creates its own FakeTensorMode here:

with FakeTensorMode(allow_non_fake_inputs=True) as fakemode:
    model = make_fx(subgraph, ...)(*fake_inputs)

Meanwhile, PyTorch's dynamo / torch.compile (used inside ComfyUI's sampler / model patching) already has its own active FakeTensorMode higher up the call stack. → Two modes clash → assertion crash.

could you share your pipeline and models ? let me try to reproduce it.

[RoyalXXX]

could you share your pipeline and models ? let me try to reproduce it.

I get this error regardless of the pipeline or models. Even in this very simple case:

[openvino-dev-samples]

Hi @RoyalXXX Have you upgraded your comfyui to latest version ? I cannot reproduce your error with following workflow:

[RoyalXXX]

Have you upgraded your comfyui to latest version ?

@openvino-dev-samples yeah, I’ve updated ComfyUI using update_comfyui.bat. Or is there anything else I need to do?

Python version: 3.13.11 (tags/v3.13.11:6278944, Dec 5 2025, 16:26:58) [MSC v.1944 64 bit (AMD64)]
ComfyUI version: 0.17.0
comfy-aimdo version: 0.2.12
comfy-kitchen version: 0.2.8
ComfyUI frontend version: 1.41.21
ComfyUI-OpenVINO 1.2.1

aiohappyeyeballs                       2.6.1
aiohttp                                3.13.3
aiosignal                              1.4.0
alembic                                1.18.4
annotated-doc                          0.0.4
annotated-types                        0.7.0
anyio                                  4.12.1
attrs                                  25.4.0
av                                     16.1.0
blake3                                 1.0.8
certifi                                2026.1.4
cffi                                   2.0.0
chardet                                7.1.0
charset-normalizer                     3.4.4
click                                  8.3.1
colorama                               0.4.6
comfy-aimdo                            0.2.12
comfy-kitchen                          0.2.8
comfyui-embedded-docs                  0.4.3
comfyui_frontend_package               1.41.21
comfyui-manager                        4.1b1
comfyui_workflow_templates             0.9.26
comfyui-workflow-templates-core        0.3.173
comfyui-workflow-templates-media-api   0.3.64
comfyui-workflow-templates-media-image 0.3.107
comfyui-workflow-templates-media-other 0.3.146
comfyui-workflow-templates-media-video 0.3.62
cryptography                           46.0.5
einops                                 0.8.2
filelock                               3.25.2
frozenlist                             1.8.0
fsspec                                 2026.2.0
gitdb                                  4.0.12
GitPython                              3.1.46
glfw                                   2.10.0
greenlet                               3.3.2
h11                                    0.16.0
hf-xet                                 1.3.0
httpcore                               1.0.9
httpx                                  0.28.1
huggingface_hub                        1.4.1
idna                                   3.11
Jinja2                                 3.1.6
kornia                                 0.8.2
kornia_rs                              0.1.10
Mako                                   1.3.10
markdown-it-py                         4.0.0
MarkupSafe                             3.0.3
mdurl                                  0.1.2
mpmath                                 1.3.0
multidict                              6.7.1
networkx                               3.6.1
numpy                                  2.3.5
openvino                               2026.0.0  20965
openvino-telemetry                     2025.2.0
packaging                              26.0
pillow                                 12.1.1
pip                                    26.0.1
propcache                              0.4.1
psutil                                 7.2.2
pycparser                              3.0
pydantic                               2.12.5
pydantic_core                          2.41.5
pydantic-settings                      2.13.1
pygit2                                 1.19.1
PyGithub                               2.8.1
Pygments                               2.19.2
PyJWT                                  2.12.0
PyNaCl                                 1.6.2
PyOpenGL                               3.1.10
PyOpenGL-accelerate                    3.1.10
python-dotenv                          1.2.1
PyYAML                                 6.0.3
regex                                  2026.2.19
requests                               2.32.5
rich                                   14.3.3
safetensors                            0.7.0
scipy                                  1.17.1
sentencepiece                          0.2.1
setuptools                             82.0.0
shellingham                            1.5.4
simpleeval                             1.0.4
smmap                                  5.0.3
spandrel                               0.4.2
SQLAlchemy                             2.0.48
sympy                                  1.14.0
tokenizers                             0.22.2
toml                                   0.10.2
torch                                  2.10.0
torchaudio                             2.10.0
torchsde                               0.2.6
torchvision                            0.25.0
tqdm                                   4.67.3
trampoline                             0.1.2
transformers                           5.3.0
typer                                  0.24.1
typer-slim                             0.24.0
typing_extensions                      4.15.0
typing-inspection                      0.4.2
urllib3                                2.6.3
uv                                     0.10.9
yarl                                   1.23.0

[okolenmi]

I have updated the node to support torch 2.10. Maybe you can help to verify, thanks a lot!

Didn't expected it so fast. Tested it. Works for me. Thank you!
I have a question about performance. Does it possible to calculate model faster on first run? It's calculated mostly on a single core of my CPU (in GPU mode, but I'm not sure it's possible to calculate it faster on GPU).

The main question is about uniform load on CPU. Only one core performing significant calculations, while other cores loaded on less than 25% in average. With better utilization of other cores the node would be much more useful as it may result in 2x+ faster initialization (depending on how much cores user have).

[okolenmi]

After some usage I have found an error, which happens when cache needs to be cleared (happens when I have ~80% VRAM used, so old models which are not used in a run would be cleared from cache).

Traceback (most recent call last):
  File "/usr/lib/python3.14/threading.py", line 1082, in _bootstrap_inner
    self._context.run(self.run)
    ~~~~~~~~~~~~~~~~~^^^^^^^^^^
  File "/usr/lib/python3.14/threading.py", line 1024, in run
    self._target(*self._args, **self._kwargs)
    ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/okolenmi/Comfy/ComfyUI/main.py", line 316, in prompt_worker
    comfy.model_management.soft_empty_cache()
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File "/home/okolenmi/Comfy/ComfyUI/comfy/model_management.py", line 1701, in soft_empty_cache
    torch.xpu.empty_cache()
    ~~~~~~~~~~~~~~~~~~~~~^^
  File "/home/okolenmi/Comfy/venv/lib/python3.14/site-packages/torch/xpu/memory.py", line 36, in empty_cache
    torch._C._xpu_emptyCache()
    ~~~~~~~~~~~~~~~~~~~~~~~~^^
RuntimeError: level_zero backend failed with error: 20 (UR_RESULT_ERROR_DEVICE_LOST)

Similar error happens even after a successful run on OpenVINO node, when I'm running a workflow without this node. Something is wrong with cache management. Not sure the error is connected to this node directly or it's a bug of XPU library. Anyway, I have never seen this error without usage of OpenVINO node.

[openvino-dev-samples]

After some usage I have found an error, which happens when cache needs to be cleared (happens when I have ~80% VRAM used, so old models which are not used in a run would be cleared from cache).

Traceback (most recent call last):
  File "/usr/lib/python3.14/threading.py", line 1082, in _bootstrap_inner
    self._context.run(self.run)
    ~~~~~~~~~~~~~~~~~^^^^^^^^^^
  File "/usr/lib/python3.14/threading.py", line 1024, in run
    self._target(*self._args, **self._kwargs)
    ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/okolenmi/Comfy/ComfyUI/main.py", line 316, in prompt_worker
    comfy.model_management.soft_empty_cache()
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File "/home/okolenmi/Comfy/ComfyUI/comfy/model_management.py", line 1701, in soft_empty_cache
    torch.xpu.empty_cache()
    ~~~~~~~~~~~~~~~~~~~~~^^
  File "/home/okolenmi/Comfy/venv/lib/python3.14/site-packages/torch/xpu/memory.py", line 36, in empty_cache
    torch._C._xpu_emptyCache()
    ~~~~~~~~~~~~~~~~~~~~~~~~^^
RuntimeError: level_zero backend failed with error: 20 (UR_RESULT_ERROR_DEVICE_LOST)

Similar error happens even after a successful run on OpenVINO node, when I'm running a workflow without this node. Something is wrong with cache management. Not sure the error is connected to this node directly or it's a bug of XPU library. Anyway, I have never seen this error without usage of OpenVINO node.

Is it happening on torch-xpu only ? could you try torch-cuda or torch-cpu ?

[openvino-dev-samples]

I have updated the node to support torch 2.10. Maybe you can help to verify, thanks a lot!

Didn't expected it so fast. Tested it. Works for me. Thank you! I have a question about performance. Does it possible to calculate model faster on first run? It's calculated mostly on a single core of my CPU (in GPU mode, but I'm not sure it's possible to calculate it faster on GPU).

The main question is about uniform load on CPU. Only one core performing significant calculations, while other cores loaded on less than 25% in average. With better utilization of other cores the node would be much more useful as it may result in 2x+ faster initialization (depending on how much cores user have).

Sorry, this part is managed by openvino runtime, so there is no chance to optimize it within this node.

[okolenmi]

I have no CUDA to test it. And it's not a thing that should be tested on RAM as it's a problem of VRAM. I think, ComfyUI have it's own cache system and this node is not synchronized with it. That seems like the most logical explanation. ComfyUI can't see VRAM usage of this node or free models loaded with it.