realy I need a NVIDIA Card

[rheros]

I finished the install, the launcher tell me Found no NVIDIA driver, I have a amd RX 6800 XT, can I use this card to run the app

[github2911]

I have the same issue as @rheros. Any guidance would be appreciated.

[EliEron]

Nvidia is the industry standard when it comes to AI, so officially speaking most AI projects only support Nvidia cards.

Some users have gotten it to work using the steps in this discussion though, so I'd give that a try. I don't have an AMD card myself so I can't vouch for whether it currently works or not.

[PyrateGFXProductions]

I have gotten my RX580 to work with SD WebUI in Win11 using DirectML via this Git, which is an installable extension for SD WebUI.

I don't think it matters but I am also using this fork of SD WebUI.

I am getting speeds of about 2s/it which is pretty fast for my 10 year old PC. However, nod.ai/SHARK is faster by about 1s/it.

[cstueckrath]

You don't need nvidia
Windows: https://gitgudblog.vercel.app/posts/stable-diffusion-amd-win10 (no torch, so many things won't work (including this webui))
Linux: See my discussion here: #756

[cstueckrath]

the link works for me, try again?

[rheros]

em, over a day, the link not works for me yet, thanks for the great wall I think,
thank you in any case

[cstueckrath]

here is the same guide as a youtube video: https://www.youtube.com/watch?v=ngVBjeu66QI

[cstueckrath]

the guide is basically this:

1. Dependencies & Preparation
   I. You'll need a recent Python 3 version installed on your machine. In this guide I'm using Python version 3.10.2.
   II. You'll also need a huggingface account as well as an API access key from the huggingface settings, to download the latest version of the Stable Diffusion model (1.4 as of writing this blog post).
   III. Make sure python and pip are defined in your PATH:

> python --version

Python 3.10.2

IV. Open your Windows Terminal or cmd (as an administrator)
V. Go to Harish Anand's Diffusers repository, clone it into a folder you've created for this little project and pip install its contents. You'll also need ftfy and scipy, so pip install those as well:

git clone https://github.com/harishanand95/diffusers.git

cd diffusers && git checkout dml && pip install -e .

pip install transformers ftfy scipy

VI. The trick that makes this work is to use the latest nightly build of Onnx. You can download it here. Make sure the downloaded version matches your python version. In my case I have to download the file ort_nightly_directml-1.13.0.dev20220901005-cp310-cp310-win_amd64.whl since I'm on python version 3.10.2. Once you've downloaded it to your project folder do a:
pip install ort_nightly_directml-1.13.0.dev20220901005-cp310-cp310-win_amd64.whl
Make sure you're not using the "dot slash" (pip install ./ort_...) relative path syntax here as this is a common mistake that prevents the pip install command from finding the local file because Windows has issues with forward slashes.

VII. Now you have to generate a Onnx version of the Stable Diffusion model. To do so run

cd ./diffusers/examples/inference

huggingface-cli login

and paste in your huggingface API token when prompted for it.

If this step fails, you probably didn't accept the terms and conditions of the Stable Diffusion model yet so head over to the model and do so before retrying the step above. Then run the ./save_onnx.py script

python ./save_onnx.py

2. Usage
   There's the file called dml_onnx.py that can be used to generate your images. You can use as a reference or edit it according to your needs. At the bottom you can change the actual prompt, the seed, the output filename, the no. of steps, the guidance scale and so on.

If you want to change the default image resolution make sure that the widths and heights you choose can be divided by 8 and follow the instructions at the bottom of the dml_onnx.py file, such as re-generating a Onnx model for that new resolution.

[cstueckrath]

when this worked for you and you got your first picture with python dml_onnx.py
you can edit dml_onnx.py like this:

import random
[...]
rand = random.randint(1, 10000000000000000000)
torch.manualseed(rand)
[...]
image.save("Dml"+str(rand)+".png")

Then run this with for /l %q in (0) do python dml_onnx.py

this will use a new seed every time for the given prompt in dml_onnx.py

[rheros]

I get some info from web,

Current State:
My device : AMD RadeonRX 6800XT
My System : Windows11 (I find that Rocm not support Windows , GOD!!! )

TODO:

1. I need to Install a ubuntu on my pc
2. install rocm
3. follow this

I will have a try tomorrow , See you

[rheros]

System prepare:

1. Install Ubuntu 20.04 https://ubuntu.com/download/alternative-downloads
2. lower kernel https://blog.csdn.net/weixin_46123670/article/details/125777787
3. reboot

Graphic preprare:

1. make sure system setting can see Graphics AMD® Radeon rx 6800 xt
   • if can't see install graphics driver https://www.amd.com/zh-hans/support/graphics/amd-radeon-6000-series/amd-radeon-6800-series/amd-radeon-rx-6800-xt this is a installer ,make driven install easy
   • other version can search from here https://www.amd.com/zh-hans/support

install rocm

(doc here) https://docs.amd.com/bundle/ROCm-Installation-Guide-v5.2.3/page/How_to_Install_ROCm.html
* sudo apt-get update
* wget https://repo.radeon.com/amdgpu-install/22.20.3/ubuntu/focal/amdgpu-install_22.20.50203-1_all.deb
* sudo apt-get install ./amdgpu-install_22.20.50203-1_all.deb
* sudo amdgpu-install --usecase=rocm
* sudo amdgpu-install --usecase=dkms

check rocm install success? use shell cmd one of blow

rocm-smi
/opt/rocm/bin/rocminfo

• if display graphics info ,congratulations you success

install conda

site https://docs.conda.io/en/latest/miniconda.html#linux-installers

install stable diffusion follow this

• https://sd-webui.github.io/stable-diffusion-webui/docs/2.linux-installation.html this is the recommend way
• I git clone the repo, and run the webui.sh under the stable-diffusion-webui folder
• remember download the sd-v1-4.ckpt mode and mv it in the stable-diffusion-webui ,webui.sh need it

install pythorch in ldm env，otherwise the webui.sh will tell you ,you don't have cuda ,

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/rocm5.2/

downlond sd-v1-4.ckpt mode

https://huggingface.co/CompVis/stable-diffusion-v-1-4-original

no module named cv2

1. enter the conda env ldm and install opencv use shell
   • conda activate ldm
   • pip install opencv-contrib-python

[rheros]

• don't run webui.sh under sudo, this will make some folder user and group to root
• my RX6800 XT can display use rocm-smi, but torch show no cuda , change the the version in environment.yaml pytorch==1.12.1

[cstueckrath]

interesting - so you didn't use the ROCm pytorch?
pip3 install --upgrade torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/rocm5.1.1

[rheros]

em,I forget witch step to install it
it should be installed and I surely have installed it

[rheros]

yes, I rm my ldm env, reinstall stable diffusion, webui.sh show err i don't have a nvidia card, yeah, I need to install
pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/rocm5.2/
thanks a lot bro