Something broke down because of one of the last updates

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

As you can see I am using one of MY OWN workflows, since it comes from MY output folder:

What was expected is it working again! Its that simple
Its this workflow (should be inside the image, otherwise I provided it in text form)

[workedBefore.txt](https://github.com/user-attachments/files/22963706/workedBefore.txt)

It is literally one of my first gens, as you see its numeration,
Problem I can no longer run the workflow due to messages saying I reached limit ( torch._dynamo hit config.recompile_limit (64)), I did not have that in my successful gen previousely.

Only noticeable change I did was:
- increase windows page file size (for another workflow I believe) from 3096 to 32k then 40ishk+ (I tried to bring it back to base afterwards and changed nothing)
- Installed ComfyUI-WanAnimatePreprocess through git clone and pip install -r req, despite the manager asking to reduce SECURITY in order to install it.. (I removed it manually, did not solve my issue, I uninstalled the libraries it brought to my install (onnx and ml_dtypes), did not solve my issue, I finally started a new comfy install, same)
- **Probably did some comfy update sometime during the last 3 days**, once or twice.
- Installed nunchaku (but the new comfy install still has the new problem, and it does not have nunchaku so..)

### Actual Behavior

MY OWN workflow does not work anymore! That's what is happening
And there IS NOT REASON for it. I am getting this:
(torch._dynamo hit config.recompile_limit (64))
I get a spike in vram usage, then comfy craches:

___
The workflow is no longer able to perform the workflow with same params that worked 2-3 days ago!
(btw I RE use the workflow as is, with same input image and same params which are: 1280x720x81, and same everything even the seed,)
I was trying to see if something could work, I tried 720x720x33 it passed, but 2-3 days I was able to to do 1280x720x81.
So the workflow is still working for smaller perfs, I mean smaller consumption (smaller video), it just would not work as 2-3 days where the video was longer)

- I spent hours trying to solve this problem, I did all the following:
- remove triton files (from both 2 paths ":\Users\\AppData\Local\Temp\torchinductor_\triton>" and "\Users\\.triton") -> restart and try the same workflow. = FAILED. OOR
- increase dynamo cache size limit from 64 to 128 -> max gpu (comfy frozen)
- Same but with reserve-ram 2 -> same (comfy not advancing at ksampler level)
- Same but reserve-ram 4 -> FAILED (error and crach)
- cache size limit from at 96 + reserve ram 2 (error: torch._dynamo hit config.recompile_limit (96))
- try ANOTHER WORKFLOW at first, then try the current workflow (in case of 2.8.0 first gen vram excessive usage) -> no effect.

C shows at high usage but it stil has 160 free GBs. Whichever page file size I experiment I never go below a resaonable free space and certainly not below 0)

### Steps to Reproduce

use the workflow on this image or txt file:

[workedBefore.txt](https://github.com/user-attachments/files/22964077/workedBefore.txt)

Have triton and sage attention (see my versions in the pip list I posted below).
Run.
(windows page file can be ANY size I get always the same error it seems: https://www.reddit.com/r/StableDiffusion/comments/1o4kntc/psa_increasing_paging_file_size_helps_preventing/)

### Debug Logs

```powershell
FETCH ComfyRegistry Data: 100/100
FETCH ComfyRegistry Data [DONE]
[ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json [DONE]
[ComfyUI-Manager] All startup tasks have been completed.
got prompt
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load devicG: cuda:0, offload devicG: cpu, dtypG: torch.bfloat16
Requested to load CLIPVisionModelProjection
loaded completely 21785.8 1208.09814453125 True
Using scaled fp8: fp8 matrix mult: False, scale input: False
CLIP/text encoder model load devicG: cuda:0, offload devicG: cpu, current: cpu, dtypG: torch.float16
Requested to load WanTEModel
loaded completely 20505.5744140625 6419.477203369141 True
Requested to load WanVAE
loaded completely 4358.422206878662 242.02829551696777 True
gguf qtypes: F16 (694), Q8_0 (400), F32 (1)
model weight dtype torch.float16, manual cast: None
model_type FLOW
lora key not loaded: diffusion_model.blocks.0.cross_attn.k_img.diff_b
lora key not loaded: diffusion_model.blocks.0.cross_attn.k_img.lora_down.weight

.....

lora key not loaded: diffusion_model.blocks.9.cross_attn.v_img.lora_up.weight
lora key not loaded: diffusion_model.img_emb.proj.0.diff
lora key not loaded: diffusion_model.img_emb.proj.0.diff_b
lora key not loaded: diffusion_model.img_emb.proj.1.diff_b
lora key not loaded: diffusion_model.img_emb.proj.1.lora_down.weight
lora key not loaded: diffusion_model.img_emb.proj.1.lora_up.weight
lora key not loaded: diffusion_model.img_emb.proj.3.diff_b
lora key not loaded: diffusion_model.img_emb.proj.3.lora_down.weight
lora key not loaded: diffusion_model.img_emb.proj.3.lora_up.weight
lora key not loaded: diffusion_model.img_emb.proj.4.diff
lora key not loaded: diffusion_model.img_emb.proj.4.diff_b
Requested to load WAN21
loaded partially 8588.889382366657 8580.986450195312 0
Attempting to release mmap (235)
Patching comfy attention to use sageattn
0%| | 0/2 [00:00...
comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.APPLY_MODEL, transformer_options)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
).execute(x, t, c_concat, c_crossattn, control, transformer_options, **kwargs)
~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\patcher_extension.py", line 113, in execute
return self.wrappers[self.idx](self, *args, **kwargs)
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy_api\torch_helpers\torch_compile.py", line 26, in apply_torch_compile_wrapper
return executor(*args, **kwargs)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\patcher_extension.py", line 105, in __call__
return new_executor.execute(*args, **kwargs)
~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\patcher_extension.py", line 112, in execute
return self.original(*args, **kwargs)
~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\model_base.py", line 200, in _apply_model
model_output = self.diffusion_model(xc, t, context=context, control=control, transformer_options=transformer_options, **extra_conds).float()
~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1773, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1784, in _call_impl
return forward_call(*args, **kwargs)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 614, in forward
return comfy.patcher_extension.WrapperExecutor.new_class_executor(
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
...<2 lines>...
comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.DIFFUSION_MODEL, transformer_options)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
).execute(x, timestep, context, clip_fea, time_dim_concat, transformer_options, **kwargs)
~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\patcher_extension.py", line 112, in execute
return self.original(*args, **kwargs)
~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 634, in _forward
return self.forward_orig(x, timestep, context, clip_fea=clip_fea, freqs=freqs, transformer_options=transformer_options, **kwargs)[:, :, :t, :h, :w]
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 579, in forward_orig
x = block(x, e=e0, freqs=freqs, context=context, context_img_len=context_img_len, transformer_options=transformer_options)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\_dynamo\eval_frame.py", line 375, in __call__
return super().__call__(*args, **kwargs)
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1773, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1784, in _call_impl
return forward_call(*args, **kwargs)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\_dynamo\eval_frame.py", line 736, in compile_wrapper
return fn(*args, **kwargs)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1773, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1784, in _call_impl
return forward_call(*args, **kwargs)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 236, in forward
torch.addcmul(repeat_e(e[0], x), self.norm1(x), 1 + repeat_e(e[1], x)),
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 235, in torch_dynamo_resume_in_forward_at_236
y = self.self_attn(
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 243, in torch_dynamo_resume_in_forward_at_235
x = x + self.cross_attn(self.norm3(x), context, context_img_len=context_img_len, transformer_options=transformer_options)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 243, in torch_dynamo_resume_in_forward_at_243
x = x + self.cross_attn(self.norm3(x), context, context_img_len=context_img_len, transformer_options=transformer_options)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 244, in torch_dynamo_resume_in_forward_at_243
y = self.ffn(torch.addcmul(repeat_e(e[3], x), self.norm2(x), 1 + repeat_e(e[4], x)))
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\ComfyUI\comfy\ldm\wan\model.py", line 244, in torch_dynamo_resume_in_forward_at_244
y = self.ffn(torch.addcmul(repeat_e(e[3], x), self.norm2(x), 1 + repeat_e(e[4], x)))
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1773, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1784, in _call_impl
return forward_call(*args, **kwargs)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\container.py", line 244, in forward
input = module(input)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1773, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1784, in _call_impl
return forward_call(*args, **kwargs)
File "G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\activation.py", line 738, in forward
return F.gelu(input, approximate=self.approximate)
~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
torch.OutOfMemoryError: Allocation on device

Got an OOM, unloading all loaded models.

G:\ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable>pause
Press any key to continue . . .

__

And my pip list currently (new comfy install has runs ONLY the workflow that I am trying to make REWORK after 2-3 days from it working fine):
ComfyUI_windows_portable_nvidia171025\ComfyUI_windows_portable\python_embeded>python.exe -m pip list
Package Version
-------------------------- ---------------------------
accelerate 1.10.1
aiofiles 24.1.0
aiohappyeyeballs 2.6.1
aiohttp 3.12.15
aiohttp_socks 0.10.1
aiosignal 1.4.0
alembic 1.16.5
annotated-types 0.7.0
attrs 25.3.0
av 15.1.0
certifi 2025.8.3
cffi 2.0.0
chardet 5.2.0
charset-normalizer 3.4.3
click 8.3.0
clip-interrogator 0.6.0
color-matcher 0.6.0
colorama 0.4.6
coloredlogs 15.0.1
comfyui-embedded-docs 0.3.0
comfyui_frontend_package 1.28.7
comfyui_workflow_templates 0.1.95
contourpy 1.3.3
cryptography 46.0.3
cycler 0.12.1
ddt 1.7.2
diffusers 0.35.2
docutils 0.22.2
einops 0.8.1
filelock 3.19.1
flatbuffers 25.9.23
fonttools 4.60.1
frozenlist 1.7.0
fsspec 2025.9.0
ftfy 6.3.1
gguf 0.17.1
gitdb 4.0.12
GitPython 3.1.45
greenlet 3.2.4
h11 0.16.0
h2 4.3.0
hpack 4.1.0
huggingface-hub 0.35.3
humanfriendly 10.0
hyperframe 6.1.0
idna 3.10
imageio 2.37.0
imageio-ffmpeg 0.6.0
importlib_metadata 8.7.0
Jinja2 3.1.6
jsonschema 4.25.1
jsonschema-specifications 2025.9.1
kiwisolver 1.4.9
kornia 0.8.1
kornia_rs 0.1.9
lark 1.3.0
Mako 1.3.10
markdown-it-py 4.0.0
MarkupSafe 3.0.3
matplotlib 3.10.7
matrix-nio 0.25.2
mdurl 0.1.2
mpmath 1.3.0
mss 10.1.0
multidict 6.6.4
networkx 3.5
numpy 2.2.6
onnxruntime 1.23.1
open_clip_torch 3.2.0
opencv-python 4.12.0.88
opencv-python-headless 4.12.0.88
packaging 25.0
peft 0.17.1
pillow 11.3.0
pip 25.2
propcache 0.3.2
protobuf 6.33.0
psutil 7.1.0
pycparser 2.23
pycryptodome 3.23.0
pydantic 2.11.9
pydantic_core 2.33.2
pydantic-settings 2.11.0
pygit2 1.18.2
PyGithub 2.8.1
Pygments 2.19.2
PyJWT 2.10.1
PyNaCl 1.6.0
pyparsing 3.2.5
pyreadline3 3.5.4
python-dateutil 2.9.0.post0
python-dotenv 1.1.1
python-socks 2.7.2
PyYAML 6.0.3
referencing 0.37.0
regex 2025.9.18
requests 2.32.5
rich 14.2.0
rpds-py 0.27.1
safetensors 0.6.2
sageattention 2.2.0+cu128torch2.8.0.post3
scipy 1.16.2
sentencepiece 0.2.1
setuptools 80.9.0
shellingham 1.5.4
six 1.17.0
smmap 5.0.2
soundfile 0.13.1
spandrel 0.4.1
SQLAlchemy 2.0.43
sympy 1.14.0
timm 1.0.20
tokenizers 0.22.1
toml 0.10.2
torch 2.8.0+cu129
torchaudio 2.8.0+cu129
torchsde 0.2.6
torchvision 0.23.0+cu129
tqdm 4.67.1
trampoline 0.1.2
transformers 4.56.2
triton-windows 3.4.0.post20
typer 0.19.2
typing_extensions 4.15.0
typing-inspection 0.4.1
unpaddedbase64 2.1.0
urllib3 2.5.0
uv 0.9.3
wcwidth 0.2.14
yarl 1.20.1
zipp 3.23.0
```

### Other

This is driving me insane, what can I do?
I am legit thinking a spyware or crypto that hides at every gen (by eating a bit of vram) could be there, but:
- doing a new comfy install
- changing browser.. etc
made believe it might not the case but who knows?
After all one of the custom nodes asked me to remove security (i mean the manager asked) for its installation, i did not listen and went installing it with git clone and git install -r req txt file, (ComfyUI-WanAnimatePreprocess) unlikely but who knows. I mean i removed it and removed the 2 libraries it installed, but maybe it did something else i did not see. The source seems reliable (kijai node I believe) so Idk...

I was intially thinking about changes in wanvideo wrapper (since I did git pulls trying to run the flash upscaler (latest 1 step workflow that upscales) and need his latest code, i had to make the custom node becomes git..table sicen it was only installed by manager initially) so I removed that ndoe later on when i discovede the workflow does not rely on it, still same problem.
I also wondered if it was a change on torchcompilemodelwanvideoV2 node (to adapt to latest torch versions) that broke something? Since I am reaching the limit torch._dynamo hit config.recompile_limit (64) or 96 faster whereas i was NEVER DOIN THAT 2-3 days ago with SAME WORKFLOW.
I want to FIX IT.
if a workflow WORKED IN THE PAST (2-3 days) I SHOULD BE ABLE TO MAKE IT WORK AGAIN right?
Thanks

Edit: Could it be a change made on comfyUI during one of the updates from 0-3 days ago, that might have affected how torchComplie node work? (Or sageattention node affected by some comfyCore change?)
Or anything else in comfyUI core or else, could not tell!