Fooocus and Nvidia 5090 / 5080 ?

[Latee77]

Great UI for beginner like i am :)

But will Fooocus work with upcoming Nvidia graphic cards too?

[mashb1t]

sure, they support all currently used technologies and the CUDA version, so i don't see an roadblock.

[Latee77]

Allright, nice, so these different card generations wont need any updates for fooocus to work :)

[theoldbar]

dont worry , by the time those cards come down to a sensible price we'll be on flux 4.0 XXXL😂

[Latee77]

:D right, lets see how those are even available in Finland after launch. All gone in second last time :/

[agapess]

Great UI for beginner like i am :)

But will Fooocus work with upcoming Nvidia graphic cards too?

iv got my 5090 , not working , lot of different error

RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

init.py:215: UserWarning:
NVIDIA GeForce RTX 5090 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90.
If you want to use the NVIDIA GeForce RTX 5090 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

[mashb1t]

same issue as in lllyasviel/stable-diffusion-webui-forge#2601

@lllyasviel turns out pytorch 2.1.0 isn't compatible with 5000 series GPUs and has to be updated.

Fooocus/launch.py

Lines 30 to 32 in 59f183a

	torch_index_url = os.environ.get('TORCH_INDEX_URL', "https://download.pytorch.org/whl/cu121")
	torch_command = os.environ.get('TORCH_COMMAND',
	f"pip install torch==2.1.0 torchvision==0.16.0 --extra-index-url {torch_index_url}")

@agapess running Fooocus using Docker after adjusting requirenents_docker.txt to more recent versions might work, but is currently not supported.

Docker quickstart: https://github.com/lllyasviel/Fooocus/blob/main/docker.md

[YudhistiraArt]

bagaimana solusi jika laptap tidak mendukung driver tersebut karena laptop versi lama

[Latee77]

Ok, some allready got 5090. If not working yet, are these kind of problems usually solved with Fooocus update itself later?

[mashb1t]

As Fooocus is currently in security support only and this is not a security feature + package updates are causing a lot of issues for all non-affected users, the chance this will get fixed soon is, realistically speaking, medium-low, but may vary depending on the relevance of the Fooocus userbase to the project owner.
I won't do the update (currently no time to handle additional coding for Fooocus), so there's that.

[Latee77]

Ok, so if i want to continue use Fooocus i need to keep my 4090 for that. And switch it to 5090 if ever get one for gaming.

Sad because Fooocus has features what havent found on web iu Forge.

Mostly because with Fooocus i can modify / fix pictures at masked area with mixing picture prompt and inpaint. Thats nice and easy way to add things what can be adjusted with image prompts settings (like colours, surface textures etc.)

[Latee77]

Is there other AI like Fooocus where is possible to modify parts of pictures by masking area, and mix it with text prompt, image prompt and inpaint? Like Fooocus advantage settings makes it possible.

Easy and fast to use, and suitable for novice users like me :)

[IPv6]

If you want free solution (not Photoshop) - then InvokeAI. It will require some youtube tutorials to watch, though.
Or Krita with krita-ai-diffusion addon

[Latee77]

Thanks, i’ll check InvokeAI, if it got change to mix masking, image prompt and inpaint. And works later with 5090 too.

[Latee77]

From other topic:
” you just need to activate the Python virtual environment where Fooocus is installed and type the pip command above (it installs the necessary PyTorch updates)”

Works with PyTorch update, all like Heprean to me, but when i some day got 5090 need to ask some advice for this update :D

[AjunNg]

Where is the python venv in Fooocus folder?

[alibakhtiari2]

Hi

I fixed the issue. visit my page. https://github.com/alibakhtiari2/fooocusrtx508090

i fixed the dependencies. now its running on rtx 50 series.

please share

[Latee77]

Still waiting for 5090 to arrive, but now Fooocus is fixed simpliest way in your link? Just download and install and all necessary fixes are done?

Thats clever thanks 👍🏻

[alibakhtiari2]

Hi. It took me 7 hours to fix the dependencies.

Rtx 50xx is coming with newer cuda version so fooocus PyTorch is not compatible.

Finally I could fix it.

Its iteration rate is 5.25 on Rtx 5080
Highest quality and default model

[Olekidh]

@alibakhtiari2 Holy smokes dude, you are a legend. Running perfectly locally on 4090 and 5090

[VulpineKitket666]

LEGEND, DROP YOU WALLET OR SMTH <3

[galactose]

To provide an update, I was able to get fooocus working on Linux Mint with an RTX5080 by following the installation steps in the README.md under Linux (Using Python Venv). With the virtual environment activated (source fooocus_env/bin/activate), in the root of the fooocus directory upgrade pytorch with pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128, then python entry_with_update.py.

[Latee77]

Finally got my 5090 and with instructions got Fooocus running. More on [lllyasviel/Fooocus] topic, but basically just needed to update pytorchs.

what pytorch model and what pyton model and to what folder, what is pip installation etc. Gemini helped a lot and guided through the whole thing.

Just a few tests with Fooocus, but does not seems to be lot faster. Maybe with heavier AI programs it benefits more.

Same with latest CoD BO6, almost the same fps with 4K as 4090, but 5090 is a lot louder and hotter xD ( yes all ROPS are there).