AssertionError: Torch not compiled with CUDA enabled - Possible to update manually?

[MaestroFury]

Hey guys,

Today Fooocus decided not to work. I'm not sure what happened, but it happened after the latest update. I tried to reinstalled Cuda (have version 12.3). I installed Anaconda and reinstalled PyTorch as that was a recommendation on google. I didn't see how it was connected since the libraries are self contained and I don't believe Fooocus is reaching out to the system's resources.

Big surprise. It didn't work.

Now i'm stuck. I think redownloading Fooocus might fix it, but man... installing it in the first place took ages, with huge files to download. Is there any way to selectively update these libraries, or why did the tool break like this? Is there some "check everything" update command I should be aware of?

Here is the full dump.

File "threading.py", line 1016, in bootstrap_inner
File "threading.py", line 953, in run
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\modules\async_worker.py", line 25, in worker
import modules.default_pipeline as pipeline
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\modules\default_pipeline.py", line 1, in
import modules.core as core
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\modules\core.py", line 1, in
from modules.patch import patch_all
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\modules\patch.py", line 6, in
import fcbh.model_base
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\backend\headless\fcbh\model_base.py", line 2, in
from fcbh.ldm.modules.diffusionmodules.openaimodel import UNetModel
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\backend\headless\fcbh\ldm\modules\diffusionmodules\openaimodel.py", line 16, in
from ..attention import SpatialTransformer
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\backend\headless\fcbh\ldm\modules\attention.py", line 10, in
from .sub_quadratic_attention import efficient_dot_product_attention
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\backend\headless\fcbh\ldm\modules\sub_quadratic_attention.py", line 27, in
from fcbh import model_management
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\backend\headless\fcbh\model_management.py", line 114, in
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\Fooocus\backend\headless\fcbh\model_management.py", line 83, in get_torch_device
return torch.device(torch.cuda.current_device())
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\python_embeded\lib\site-packages\torch\cuda_init.py", line 674, in current_device
lazy_init()
File "C:\Users\Administrator\Desktop\Fooocus_win64_2-1-791\python_embeded\lib\site-packages\torch\cuda_init.py", line 239, in _lazy_init
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

[MaestroFury]

Solved it.

my run.bat file had this text in it.

.\python_embeded\python.exe -s Fooocus\entry_with_update.py
pause

.\python_embeded\python.exe -m pip uninstall torch torchvision torchaudio torchtext functorch xformers -y

.\python_embeded\python.exe -m pip install torch-directml

.\python_embeded\python.exe -s Fooocus\entry_with_update.py -directml
pause

I don't know how it ended up like this. I didn't touch the bat file since installing. But removing ALL this, and just putting in

.\python_embeded\python.exe -s Fooocus\entry_with_update.py --directml
fixed the problem.

EDIT: spoke too soon. Now I am getting

RuntimeError: Device type privateuseone is not supported for torch.Generator() api

when I actually try to generate something.

[lllyasviel]

you are likely mislead by some bad online tutorials and broken your OS.
fooocus does not need you install any environment like cuda.
you may try download fooocus again and just run it directly.
please only follow official instructions.

[MaestroFury]

You are right. I just re-downloaded foocus, and copy pasted all the models so I didn't have to redownload everything. Works perfect. Now I know I also don't have to redownload everything.