Comfyui manager "update all" seem to enforce nvidia package on AMD hardware

### Custom Node Testing

- [ ] 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

hello,
Could you check if your last you updates run on amd cards ?
I can't use the "update all" function in comfyui manager anymore...
if your runing an AMD card (pytorch ROCm package) ; the updater remove it and replace it with a normal pytorch package which is meant for nvidia card.
This completly brake comfyui than can't be launched anymore until you reinstall the proper package.

**This what I expect to have at the start**
Command: ['/home/zeuss194/COMFY/ComfyUI/Comfyenv313/bin/python', 'main.py']
[START] Security scan
[ComfyUI-Manager] Using `uv` as Python module for pip operations.
Using Python 3.13.11 environment at: Comfyenv313
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2025-12-20 22:18:29.004
** Platform: Linux
** Python version: 3.13.11 (main, Dec 6 2025, 08:52:52) [GCC 13.3.0]
** Python executable: /home/zeuss194/COMFY/ComfyUI/Comfyenv313/bin/python
** ComfyUI Path: /home/zeuss194/COMFY/ComfyUI
** ComfyUI Base Folder Path: /home/zeuss194/COMFY/ComfyUI
** User directory: /home/zeuss194/COMFY/ComfyUI/user
** ComfyUI-Manager config path: /home/zeuss194/COMFY/ComfyUI/user/__manager/config.ini
** Log path: /home/zeuss194/COMFY/ComfyUI/user/comfyui.log
Using Python 3.13.11 environment at: Comfyenv313
Using Python 3.13.11 environment at: Comfyenv313

Prestartup times for custom nodes:
0.0 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/rgthree-comfy
0.0 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/ComfyUI-Easy-Use
0.9 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/comfyui-manager

Checkpoint files will always be loaded safely.
Total VRAM 32624 MB, total RAM 63886 MB
pytorch version: 2.11.0.dev20251218+rocm7.1
Set: torch.backends.cudnn.enabled = False for better AMD performance.
AMD arch: gfx1201
ROCm version: (7, 1)
Set vram state to: NORMAL_VRAM
Device: cuda:0 AMD Radeon AI PRO R9700 : native
Enabled pinned memory 60691.0
Using pytorch attention
Python version: 3.13.11 (main, Dec 6 2025, 08:52:52) [GCC 13.3.0]
ComfyUI version: 0.5.1
ComfyUI frontend version: 1.34.9
[Prompt Server] web root: /home/zeuss194/COMFY/ComfyUI/Comfyenv313/lib/python3.13/site-packages/comfyui_frontend_package/static
Total VRAM 32624 MB, total RAM 63886 MB
pytorch version: 2.11.0.dev20251218+rocm7.1
Set: torch.backends.cudnn.enabled = False for better AMD performance.
AMD arch: gfx1201
ROCm version: (7, 1)
Set vram state to: NORMAL_VRAM
Device: cuda:0 AMD Radeon AI PRO R9700 : native
Enabled pinned memory 60691.0
### Loading: ComfyUI-Impact-Pack (V8.28)
[Impact Pack] Wildcard total size (0.00 MB) is within cache limit (50.00 MB). Using full cache mode.

**This what the updater does**

## ComfyUI-Manager: EXECUTE =>
['/home/zeuss194/COMFY/ComfyUI/Comfyenv313/bin/python', '-m', 'uv', 'pip',
'install', 'vllm>=0.12.0']
[!] Using Python 3.13.11 environment at: /home/zeuss194/COMFY/ComfyUI/Comfyenv313
[!] Resolved 162 packages in 540ms
[!] Uninstalled 3 packages in 712ms
[!] Installed 3 packages in 248ms
[!] - torch==2.11.0.dev20251218+rocm7.1
[!] + torch==2.9.0
[!] - torchaudio==2.10.0.dev20251219+rocm7.1
[!] + torchaudio==2.9.0
[!] - torchvision==0.25.0.dev20251219+rocm7.1
[!] + torchvision==0.24.0
[ComfyUI-Manager] restore PyTorch to nightly version
Updating: ComfyUI_MIGraphX
[ComfyUI-Manager] Queued works are completed.
{'update-comfyui': 1, 'update-main': 33}

### Actual Behavior

**This what i've got once you do a "update all **

(Comfyenv313) zeuss194@zeuss194-ThinkStation-P520:~/COMFY/ComfyUI$ python main.py
[START] Security scan
[ComfyUI-Manager] Using `uv` as Python module for pip operations.
Using Python 3.13.11 environment at: Comfyenv313
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2025-12-21 03:02:16.293
** Platform: Linux
** Python version: 3.13.11 (main, Dec 6 2025, 08:52:52) [GCC 13.3.0]
** Python executable: /home/zeuss194/COMFY/ComfyUI/Comfyenv313/bin/python
** ComfyUI Path: /home/zeuss194/COMFY/ComfyUI
** ComfyUI Base Folder Path: /home/zeuss194/COMFY/ComfyUI
** User directory: /home/zeuss194/COMFY/ComfyUI/user
** ComfyUI-Manager config path: /home/zeuss194/COMFY/ComfyUI/user/__manager/config.ini
** Log path: /home/zeuss194/COMFY/ComfyUI/user/comfyui.log
Using Python 3.13.11 environment at: Comfyenv313
Using Python 3.13.11 environment at: Comfyenv313

Prestartup times for custom nodes:
0.0 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/rgthree-comfy
0.0 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/ComfyUI-Easy-Use
0.5 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/comfyui-manager

/home/zeuss194/COMFY/ComfyUI/Comfyenv313/lib/python3.13/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.
import pynvml # type: ignore[import]
Checkpoint files will always be loaded safely.
Traceback (most recent call last):
File "/home/zeuss194/COMFY/ComfyUI/main.py", line 177, in
import execution
File "/home/zeuss194/COMFY/ComfyUI/execution.py", line 15, in
import comfy.model_management
File "/home/zeuss194/COMFY/ComfyUI/comfy/model_management.py", line 239, in
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
~~~~~~~~~~~~~~~~^^
File "/home/zeuss194/COMFY/ComfyUI/comfy/model_management.py", line 189, in get_torch_device
return torch.device(torch.cuda.current_device())
~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/home/zeuss194/COMFY/ComfyUI/Comfyenv313/lib/python3.13/site-packages/torch/cuda/__init__.py", line 1069, in current_device
_lazy_init()
~~~~~~~~~~^^
File "/home/zeuss194/COMFY/ComfyUI/Comfyenv313/lib/python3.13/site-packages/torch/cuda/__init__.py", line 410, in _lazy_init
torch._C._cuda_init()
~~~~~~~~~~~~~~~~~~~^^
RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx
[W1221 03:02:17.432260405 AllocatorConfig.cpp:28] Warning: PYTORCH_CUDA_ALLOC_CONF is deprecated, use PYTORCH_ALLOC_CONF instead (function operator())
(Comfyenv313) zeuss194@zeuss194-ThinkStation-P520:~/COMFY/ComfyUI$

### Steps to Reproduce

Comfyui in nightly mode and use "update all" with an AMD card on your pc

### Debug Logs

```powershell
Comfyenv313) zeuss194@zeuss194-ThinkStation-P520:~/COMFY/ComfyUI$ python main.py
[START] Security scan
[ComfyUI-Manager] Using `uv` as Python module for pip operations.
Using Python 3.13.11 environment at: Comfyenv313
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2025-12-21 03:02:16.293
** Platform: Linux
** Python version: 3.13.11 (main, Dec 6 2025, 08:52:52) [GCC 13.3.0]
** Python executable: /home/zeuss194/COMFY/ComfyUI/Comfyenv313/bin/python
** ComfyUI Path: /home/zeuss194/COMFY/ComfyUI
** ComfyUI Base Folder Path: /home/zeuss194/COMFY/ComfyUI
** User directory: /home/zeuss194/COMFY/ComfyUI/user
** ComfyUI-Manager config path: /home/zeuss194/COMFY/ComfyUI/user/__manager/config.ini
** Log path: /home/zeuss194/COMFY/ComfyUI/user/comfyui.log
Using Python 3.13.11 environment at: Comfyenv313
Using Python 3.13.11 environment at: Comfyenv313

Prestartup times for custom nodes:
0.0 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/rgthree-comfy
0.0 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/ComfyUI-Easy-Use
0.5 seconds: /home/zeuss194/COMFY/ComfyUI/custom_nodes/comfyui-manager

/home/zeuss194/COMFY/ComfyUI/Comfyenv313/lib/python3.13/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.
import pynvml # type: ignore[import]
Checkpoint files will always be loaded safely.
Traceback (most recent call last):
File "/home/zeuss194/COMFY/ComfyUI/main.py", line 177, in
import execution
File "/home/zeuss194/COMFY/ComfyUI/execution.py", line 15, in
import comfy.model_management
File "/home/zeuss194/COMFY/ComfyUI/comfy/model_management.py", line 239, in
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
~~~~~~~~~~~~~~~~^^
File "/home/zeuss194/COMFY/ComfyUI/comfy/model_management.py", line 189, in get_torch_device
return torch.device(torch.cuda.current_device())
~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/home/zeuss194/COMFY/ComfyUI/Comfyenv313/lib/python3.13/site-packages/torch/cuda/__init__.py", line 1069, in current_device
_lazy_init()
~~~~~~~~~~^^
File "/home/zeuss194/COMFY/ComfyUI/Comfyenv313/lib/python3.13/site-packages/torch/cuda/__init__.py", line 410, in _lazy_init
torch._C._cuda_init()
~~~~~~~~~~~~~~~~~~~^^
RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx
[W1221 03:02:17.432260405 AllocatorConfig.cpp:28] Warning: PYTORCH_CUDA_ALLOC_CONF is deprecated, use PYTORCH_ALLOC_CONF instead (function operator())
(Comfyenv313) zeuss194@zeuss194-ThinkStation-P520:~/COMFY/ComfyUI$
```

### Other

_No response_