triton + sageattention error: RuntimeError: PassManager::run failed

### Expected Behavior

Expected image gen to start as intended

### Actual Behavior

Image gen crashed on 1st step.

### Steps to Reproduce

added --use-sage-attention tag
made a default workflow

### Debug Logs

```powershell
loaded completely 9.5367431640625e+25 4897.0483474731445 True
0%| | 0/40 [00:00}>>'
0%| | 0/40 [00:02
stages["ttgir"] = lambda src, metadata: self.make_ttgir(src, metadata, options, self.capability)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\cui\python_embeded\Lib\site-packages\triton\backends\nvidia\compiler.py", line 195, in make_ttgir
pm.run(mod)
RuntimeError: PassManager::run failed

Prompt executed in 0.12 seconds
```

### Other

GPU: Nvidia GTX1080
Steps:
1. Installed Python 3.11
2. Installed torch 2.5.1
3. Installed triton-3.1.0-cp311 via ComfyUI manager
4. Installed sageattention-1.0.6 via ComfyUI manager