不支持cuda的电脑上运行报错

本地电脑没有cuda，使用以下命令在conda虚拟环境中单独安装pytorch后再启动conflowyspace软件安装：
```
conda install pytorch torchvision torchaudio cpuonly -c pytorch
```
在安装完后启动软件，提示
```
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled
```
等好长一段时间后报timeout后能进入到界面，但是不能正常使用。
查看conflowyspace的日志，执行的命令是
```
anaconda\comflowy\python.exe main.py --enable-cors-header
```
我在cmd里面，进入comflUI文件夹后单独执行
```
conda activate comflowy
python main.py --disable-cuda-malloc --use-split-cross-attention --cpu
```
日志如下
```
Total VRAM 14255 MB, total RAM 14255 MB
pytorch version: 2.3.0
Set vram state to: DISABLED
Device: cpu
Using split optimization for cross attention
[Prompt Server] web root: C:\comflowy\ComfyUI\ComfyUI\web

Import times for custom nodes:
0.0 seconds: C:\comflowy\ComfyUI\ComfyUI\custom_nodes\websocket_image_save.py

Starting server

To see the GUI go to: http://127.0.0.1:8188
```
界面能正常打开
是否能支持没有cuda的电脑运行conflowyspace

The issue is about Comflowyspace failing on non-CUDA systems despite a CPU-only PyTorch install. Look at the main.py entry point and the startup logic that checks for CUDA. The user found a working command line with --cpu flag; compare how the software launches (anaconda\comflowy\python.exe main.py --enable-cors-header) versus the manual command. Check for hardcoded CUDA assertions or missing CPU fallback in the launch script. The fix likely involves modifying the startup script to pass --cpu or similar flags when CUDA is unavailable.

Written by the indexing model from the issue text.