Latest version won't start - AttributeError: module 'triton' has no attribute 'language'

### 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

Start up normally

### Actual Behavior

Previous version working fine... just upgraded to latest version and now get failure on startup. I have no idea what it means.

### Steps to Reproduce

Update to latest version in Stability Matrix on PC (Windows 11)

### Debug Logs

```powershell
Adding extra search path checkpoints F:\Program\StabilityMatrix\Models\StableDiffusion
Adding extra search path vae F:\Program\StabilityMatrix\Models\VAE
Adding extra search path loras F:\Program\StabilityMatrix\Models\Lora
Adding extra search path loras F:\Program\StabilityMatrix\Models\LyCORIS
Adding extra search path upscale_models F:\Program\StabilityMatrix\Models\ESRGAN
Adding extra search path upscale_models F:\Program\StabilityMatrix\Models\RealESRGAN
Adding extra search path upscale_models F:\Program\StabilityMatrix\Models\SwinIR
Adding extra search path embeddings F:\Program\StabilityMatrix\Models\Embeddings
Adding extra search path hypernetworks F:\Program\StabilityMatrix\Models\Hypernetwork
Adding extra search path controlnet F:\Program\StabilityMatrix\Models\ControlNet
Adding extra search path controlnet F:\Program\StabilityMatrix\Models\T2IAdapter
Adding extra search path clip F:\Program\StabilityMatrix\Models\TextEncoders
Adding extra search path clip_vision F:\Program\StabilityMatrix\Models\ClipVision
Adding extra search path diffusers F:\Program\StabilityMatrix\Models\Diffusers
Adding extra search path gligen F:\Program\StabilityMatrix\Models\GLIGEN
Adding extra search path vae_approx F:\Program\StabilityMatrix\Models\ApproxVAE
Adding extra search path ipadapter F:\Program\StabilityMatrix\Models\IpAdapter
Adding extra search path ipadapter F:\Program\StabilityMatrix\Models\IpAdapters15
Adding extra search path ipadapter F:\Program\StabilityMatrix\Models\IpAdaptersXl
Adding extra search path prompt_expansion F:\Program\StabilityMatrix\Models\PromptExpansion
Adding extra search path ultralytics F:\Program\StabilityMatrix\Models\Ultralytics
Adding extra search path ultralytics_bbox F:\Program\StabilityMatrix\Models\Ultralytics\bbox
Adding extra search path ultralytics_segm F:\Program\StabilityMatrix\Models\Ultralytics\segm
Adding extra search path sams F:\Program\StabilityMatrix\Models\Sams
Adding extra search path diffusion_models F:\Program\StabilityMatrix\Models\DiffusionModels
Checkpoint files will always be loaded safely.
Total VRAM 16375 MB, total RAM 65443 MB
pytorch version: 2.9.0+cu128
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 4080 : cudaMallocAsync
working around nvidia conv3d memory bug.
Traceback (most recent call last):
File "F:\Program\StabilityMatrix\Packages\ComfyUI\main.py", line 149, in
import execution
File "F:\Program\StabilityMatrix\Packages\ComfyUI\execution.py", line 16, in
import nodes
File "F:\Program\StabilityMatrix\Packages\ComfyUI\nodes.py", line 24, in
import comfy.diffusers_load
File "F:\Program\StabilityMatrix\Packages\ComfyUI\comfy\diffusers_load.py", line 3, in
import comfy.sd
File "F:\Program\StabilityMatrix\Packages\ComfyUI\comfy\sd.py", line 9, in
from .ldm.models.autoencoder import AutoencoderKL, AutoencodingEngine
File "F:\Program\StabilityMatrix\Packages\ComfyUI\comfy\ldm\models\autoencoder.py", line 11, in
import comfy.ops
File "F:\Program\StabilityMatrix\Packages\ComfyUI\comfy\ops.py", line 73, in
@torch.compiler.disable()
File "F:\Program\StabilityMatrix\Packages\ComfyUI\venv\lib\site-packages\torch\compiler\__init__.py", line 253, in disable
import torch._dynamo
File "F:\Program\StabilityMatrix\Packages\ComfyUI\venv\lib\site-packages\torch\_dynamo\__init__.py", line 13, in
from . import (
File "F:\Program\StabilityMatrix\Packages\ComfyUI\venv\lib\site-packages\torch\_dynamo\aot_compile.py", line 15, in
from . import convert_frame
File "F:\Program\StabilityMatrix\Packages\ComfyUI\venv\lib\site-packages\torch\_dynamo\convert_frame.py", line 57, in
from torch._dynamo.symbolic_convert import TensorifyState
File "F:\Program\StabilityMatrix\Packages\ComfyUI\venv\lib\site-packages\torch\_dynamo\symbolic_convert.py", line 53, in
from torch._dynamo.exc import ObservedException, TensorifyScalarRestartAnalysis
File "F:\Program\StabilityMatrix\Packages\ComfyUI\venv\lib\site-packages\torch\_dynamo\exc.py", line 45, in
from .utils import counters
File "F:\Program\StabilityMatrix\Packages\ComfyUI\venv\lib\site-packages\torch\_dynamo\utils.py", line 2417, in
common_constant_types.add(triton.language.dtype)
AttributeError: module 'triton' has no attribute 'language'
[W1023 22:30:23.000000000 AllocatorConfig.cpp:28] Warning: PYTORCH_CUDA_ALLOC_CONF is deprecated, use PYTORCH_ALLOC_CONF instead (function operator ())
```

### Other

_No response_