RuntimeError: Your device does not support the current version of Torch/CUDA! after installing adetailer.

Nobody has claimed this yet.

```
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-527-g720b80da
Commit hash: 720b80daea9070aa7896d8ce2c4a6cf8daf927cf
Traceback (most recent call last):
File "C:\AI\webui_forge_cu124_torch24\webui\launch.py", line 54, in
main()
File "C:\AI\webui_forge_cu124_torch24\webui\launch.py", line 42, in main
prepare_environment()
File "C:\AI\webui_forge_cu124_torch24\webui\modules\launch_utils.py", line 434, in prepare_environment
raise RuntimeError(
RuntimeError: Your device does not support the current version of Torch/CUDA! Consider download another version:
```

Using the prebuilt CUDA 12.4 + Pytorch 2.4 package
The program runs as expected, closes, reopens with no issues. I tested installing vectorscope CC, no issues. The moment I install adetailer, forge won't launch. Removing the extension doesn't fix the problem. I suspect that there might be some instability with how the prepackaged torch works that breaks its compatibility with adetailer and possibly other extensions.

My guess is this part of the installation is causing issues with the prepackaged install:

```
Installing collected packages: py-cpuinfo, torch, ultralytics-thop, torchvision, seaborn, ultralytics
Attempting uninstall: torch
Found existing installation: torch 2.4.0+cu124
Uninstalling torch-2.4.0+cu124:
Successfully uninstalled torch-2.4.0+cu124
Attempting uninstall: torchvision
Found existing installation: torchvision 0.19.0+cu124
Uninstalling torchvision-0.19.0+cu124:
Successfully uninstalled torchvision-0.19.0+cu124
Successfully installed py-cpuinfo-9.0.0 seaborn-0.13.2 torch-2.4.1 torchvision-0.19.1 ultralytics-8.2.91 ultralytics-thop-2.0.6
```

Start with the installation traceback and the dependency versions reported for adetailer, then inspect launch.py and modules/launch_utils.py around the environment preparation check. Reproduce the prebuilt CUDA 12.4/PyTorch 2.4 setup, install adetailer, and compare the resulting torch and torchvision packages. Done means Forge launches successfully after the extension is installed without breaking the packaged runtime.

Written by the indexing model from the issue text.