Home Creators Posts Import Register Favorites Logout

Downloads

Content

SD Forge Web UI latest version with blazing fast pre-compiled Flash Attention, Sage Attention and xFormers; installers support RTX 5000 series and older GPUs.

Note: this is an older post kept for reference. The current, maintained version is Stable Diffusion WebUI Forge NEO: 1-Click Installer for Windows & Cloud.

Automatic installers for Windows, RunPod and Massed Compute with Torch 2.8, CUDA 12.9 and Flash Attention 2.8.3, and SD 1.5 models work since the November 2025 update. For more features the post points to the newer Forge NEO installer.

1-Click to install latest SD Forge Web UI on Windows, RunPod and Massed Compute and supports RTX 5000 series GPUs too with accurate Torch version and pre compiled xFormers, Sage Attention and Flash Attention

Patreon exclusive posts index to find our scripts easily, Patreon scripts updates history to see which updates arrived to which scripts and amazing Patreon special generative scripts list that you can use in any of your task.

Join discord to get help, chat, discuss and also tell me your discord username to get your special rank : SECourses Discord

Please also Star, Watch and Fork our Stable Diffusion & Generative AI  GitHub repository and join our Reddit subreddit and follow me on LinkedIn (my real profile)

=======

Latest zip file : Forge_Installer_v12.zip

27 November 2025 Update

  • Now app works with SD 1.5 models as well

  • Previously SD 1.5 models were generating black images

24 September 2025 Update

Here more updated Forge Web UI With more features : https://www.patreon.com/posts/138694680

  • Also classic Forge Web UI libraries upgraded to Torch 2.8, CUDA 12.9, Flash Attention 2.8.3 and xFormers 0.0.33 with support of literally all GPUs starting from RTX 1000 series to 5000 series including server GPUs like H100, B200, A6000 etc

8 August 2025 Update

  • Installers completey updated and now directly supports RTX 5000 series GPUs too both on RunPod, Massed Compute and Windows

    • Moreover, now we install pre-compiled Flash Attention, Sage Attention and xFormers too and speed is excellent

    • Installation is also now way faster and more robust

  • Our The Ultimate Image and Video AI Models Downloader app now supports SD Forge Web UI folder format as well so you can directly download models into accurate folders and you can download SwarmUI bundles like FLUX Bundle > https://www.patreon.com/posts/114517862

  • Now it will auto open the app in browser

  • Our SD Forge Web UI repo updated with latest updates from main repo : https://github.com/lllyasviel/stable-diffusion-webui-forge

  • Edit Windows_Start_Forge_Web_UI and modify below line to add command line start args if you know what you are doing - skip-requirements is default for Windows bat file

    • set COMMANDLINE_ARGS=--skip-requirements

    • e.g. set COMMANDLINE_ARGS=--always-normal-vram --disable-xformers

Windows Requirements

  • Python 3.10, FFmpeg, CUDA 12.8, cuDNN 9.7 or above, C++ tools and Git

  • If it doesn't work make sure to below tutorial and install everything exactly as shown in this below tutorial

  • https://youtu.be/DrhUHnYfwC0

    • Requirements are important to have CUDA, C++ Tools, MSVC

Massed Compute (Recommend Cloud) :

RunPod (Cloud):

Forge command line args can be found below check both links if you need :

 

=======

FLUX Sampler Info from a user

-Install the latest version (v12, this one). -go to your webui folder. in the address bar of your explorer type CMD, to open a console in that location. -copy and apply this command: git reset --hard f53307881bfd824dbdce6ac0d4bba04d9a74ab36 After that, you will get an error. So delete the VENV folder and install again Forge from the .bat that Furkan uploaded. (install it in the same location). Now start Forge from the webui-user.bat and there you have. Blazing fast Forge WITH Flux Samplers. NEVER UPDATE AGAIN :)

1-Click to install latest SD Forge Web UI on Windows, RunPod and Massed Compute and supports RTX 5000 series GPUs too with accurate Torch version and pre compiled xFormers, Sage Attention and Flash Attention

Patreon exclusive posts index to find our scripts easily, Patreon scripts updates history to see which updates arrived to which scripts and amazing Patreon special generative scripts list that you can use in any of your task.

Join discord to get help, chat, discuss and also tell me your discord username to get your special rank : SECourses Discord

Please also Star, Watch and Fork our Stable Diffusion & Generative AI  GitHub repository and join our Reddit subreddit and follow me on LinkedIn (my real profile)

=======

Latest zip file : Forge_Installer_v12.zip

24 September 2025 Update

Here more updated Forge Web UI With more features : https://www.patreon.com/posts/138694680

  • Also classic Forge Web UI libraries upgraded to Torch 2.8, CUDA 12.9, Flash Attention 2.8.3 and xFormers 0.0.33 with support of literally all GPUs starting from RTX 1000 series to 5000 series including server GPUs like H100, B200, A6000 etc

8 August 2025 Update

  • Installers completey updated and now directly supports RTX 5000 series GPUs too both on RunPod, Massed Compute and Windows

    • Moreover, now we install pre-compiled Flash Attention, Sage Attention and xFormers too and speed is excellent

    • Installation is also now way faster and more robust

  • Our The Ultimate Image and Video AI Models Downloader app now supports SD Forge Web UI folder format as well so you can directly download models into accurate folders and you can download SwarmUI bundles like FLUX Bundle > https://www.patreon.com/posts/114517862

  • Now it will auto open the app in browser

  • Our SD Forge Web UI repo updated with latest updates from main repo : https://github.com/lllyasviel/stable-diffusion-webui-forge

  • Edit Windows_Start_Forge_Web_UI and modify below line to add command line start args if you know what you are doing - skip-requirements is default for Windows bat file

    • set COMMANDLINE_ARGS=--skip-requirements

    • e.g. set COMMANDLINE_ARGS=--always-normal-vram --disable-xformers

Windows Requirements

  • Python 3.10, FFmpeg, CUDA 12.8, cuDNN 9.7 or above, C++ tools and Git

  • If it doesn't work make sure to below tutorial and install everything exactly as shown in this below tutorial

  • https://youtu.be/DrhUHnYfwC0

    • Requirements are important to have CUDA, C++ Tools, MSVC

Massed Compute (Recommend Cloud) :

RunPod (Cloud):

Forge command line args can be found below check both links if you need :

 

=======

FLUX Sampler Info from a user

-Install the latest version (v12, this one). -go to your webui folder. in the address bar of your explorer type CMD, to open a console in that location. -copy and apply this command: git reset --hard f53307881bfd824dbdce6ac0d4bba04d9a74ab36 After that, you will get an error. So delete the VENV folder and install again Forge from the .bat that Furkan uploaded. (install it in the same location). Now start Forge from the webui-user.bat and there you have. Blazing fast Forge WITH Flux Samplers. NEVER UPDATE AGAIN :)

1-Click to install latest SD Forge Web UI on Windows, RunPod and Massed Compute and supports RTX 5000 series GPUs too with accurate Torch version and pre compiled xFormers, Sage Attention and Flash Attention

Patreon exclusive posts index to find our scripts easily, Patreon scripts updates history to see which updates arrived to which scripts and amazing Patreon special generative scripts list that you can use in any of your task.

Join discord to get help, chat, discuss and also tell me your discord username to get your special rank : SECourses Discord

Please also Star, Watch and Fork our Stable Diffusion & Generative AI  GitHub repository and join our Reddit subreddit and follow me on LinkedIn (my real profile)

=======

Latest zip file : Forge_Installer_v12.zip

8 August 2025 Update

  • Installers completey updated and now directly supports RTX 5000 series GPUs too both on RunPod, Massed Compute and Windows

    • Moreover, now we install pre-compiled Flash Attention, Sage Attention and xFormers too and speed is excellent

    • Installation is also now way faster and more robust

  • Our The Ultimate Image and Video AI Models Downloader app now supports SD Forge Web UI folder format as well so you can directly download models into accurate folders and you can download SwarmUI bundles like FLUX Bundle > https://www.patreon.com/posts/114517862

  • Now it will auto open the app in browser

  • Our SD Forge Web UI repo updated with latest updates from main repo : https://github.com/lllyasviel/stable-diffusion-webui-forge

  • Edit Windows_Start_Forge_Web_UI and modify below line to add command line start args if you know what you are doing - skip-requirements is default for Windows bat file

    • set COMMANDLINE_ARGS=--skip-requirements

    • e.g. set COMMANDLINE_ARGS=--always-normal-vram --disable-xformers

Windows Requirements

  • Python 3.10, FFmpeg, CUDA 12.8, cuDNN 9.7 or above, C++ tools and Git

  • If it doesn't work make sure to below tutorial and install everything exactly as shown in this below tutorial

  • https://youtu.be/DrhUHnYfwC0

    • Requirements are important to have CUDA, C++ Tools, MSVC

Massed Compute (Recommend Cloud) :

RunPod (Cloud):

Forge command line args can be found below check both links if you need :

=======

FLUX Sampler Info from a user

-Install the latest version (v12, this one). -go to your webui folder. in the address bar of your explorer type CMD, to open a console in that location. -copy and apply this command: git reset --hard f53307881bfd824dbdce6ac0d4bba04d9a74ab36 After that, you will get an error. So delete the VENV folder and install again Forge from the .bat that Furkan uploaded. (install it in the same location). Now start Forge from the webui-user.bat and there you have. Blazing fast Forge WITH Flux Samplers. NEVER UPDATE AGAIN :)

Related posts

All posts in this topic: AI Image Generation and Editing Apps · New to SECourses? Start Here: Best and Most Popular SECourses Posts

Files

Comments

SECourses

thank you so much fixed with v4

Art

hi, Doctor... what of the linked arguments improve generation speed among the arguments linked? (I mean for sdxl generation pirposes)...

jccc

Can Forge UI use FLUX controlnet now?

Neil Rhodes

"AssertionError: You do not have CLIP state dict! Time taken: 0.3 sec." what am i doing wrong?

SECourses

did you see screenshot? i have selected clip. you need to select it

SECourses

i think not supported yet but you can try

SECourses

i dont know there are just too many you need to test :D

Neil Rhodes

im using the model number 10 from your downloads as a i really had no idea which si best , If i select Clip in the menu and generate at first i get a huge page of errors if i try the other clip "g" or "l" neither work i get blue screen of death and my PC reboots, PC ran into a problem and reboots, 3080 GPU 12 GB vram and 5950X CPU

SECourses

blue screen means your pc is not stable. could be RAM memory frequency and voltage

Hockey

In the picture of the car you have outlined settings. Are those the best sampler + scheduler etc. settings for Forge UI?

SECourses

nope they were just example. i would say compare

Satish Tomer

Hello Sir, where can I download Face Fusion 3

The90sPope

All i get is a black image with green dots, no matter what settings on sampler and scheduler with FLUX

SECourses

ping me from discord i can show accurate setup quickly

Mikel

I use Swarm UI, I just wish I could remember the workflows. I end up having to re-watch videos to remember where and how to do things.

Hockey

Thank you so much for the update, SD Forge has used Torch 2.3.1 for so long now. I noticed I still have Torch 2.3.1 after install. Does "2_Make_It_Work_With_RTX_5000_Series" only work with RTX 5000 Series?

SECourses

it works on all gpus. especially 3000 and above even better

Hockey

I'm going crazy, when I start Forge I get these messages: LINK : fatal error LNK1181: cannot open input file 'aio.lib' test.c LINK : fatal error LNK1181: cannot open input file 'cufile.lib' Apparently they are Cuda files but I can't find them anywhere on my PC or on the internet. Does anyone have them or know where to find them? I have looked in Cuda Samples files and downloaded Cuda 12.8 and 12.1 again but they don't show up.

SECourses

that is not an error. i get it at every application. it is linux related error ignore it

Hoàng Giang Sơn Trương

please help me. "RuntimeError: Couldn't fetch assets. Command: "git" -C "D:\AI\Forge_Installer_v9\stable-diffusion-webui-forge\repositories\stable-diffusion-webui-assets" fetch --refetch --no-auto-gc Error code: 128"

Hoàng Giang Sơn Trương

I have installed your other 1-click install versions and they run very smoothly; only this version has an error.

SECourses

well i dont know your error but it looks like git error. it could be related to your internet as well. perhaps turn off VPN try, restart pc try, and try with WARP VPN installed but it is not an error related to my installer it could be related to stable-diffusion-webui-forge itself as well

Josh S

Just wanted to say thank you very much for all the time and effort you put into this! Just bought a 5090 and was having trouble. Is there any optimizations that can help improve speed with all this extra GPU memory?

SECourses

yes it has some vram parameters. you can test each parameter and see which one works best you have to test and compare each case sadly --always-gpu Forces the application to always use GPU processing, regardless of available VRAM or system conditions. --always-high-vram Forces high VRAM usage mode, typically for systems with 12GB+ VRAM that can handle large models and batch sizes without optimization. --always-normal-vram Forces normal VRAM usage mode, suitable for mid-range GPUs with 6-12GB VRAM, balancing performance and memory usage. --always-low-vram Forces low VRAM usage mode, optimized for GPUs with 4-6GB VRAM using techniques like model offloading and reduced precision.

Hockey

I noticed that there have been some updates in the lllyasviel/stable-diffusion-webui-forge directory, however not in your version which it installs from. Are you planning to update it to match the lllyasviel/stable-diffusion-webui-forge directory? Two of them is "support for model with fp8 quantification with scaled weight" and "this PR add the ability to use GGUF model with BF16 layers" which would make the GGUF version of Flux Krea work. Also the Flux Realistic samplers have been removed but can be reinstated according to these instructions: "Before performing the procedures below, close your FORGE Go to your webui directory and enter it Right click anywhere inside the webui directory In the menu that appears, click on Open in Terminal (Now you are inside your webui to make the change in a super simple and quick way.) In the terminal that opens, only paste this command below: git reset --hard f53307881bfd824dbdce6ac0d4bba04d9a74ab36 DONE, the magic has been DONE" However the above doesn't work anymore: "WARNING:bitsandbytes.cextension:Could not find the bitsandbytes CUDA binary at WindowsPath('C:/AI/SD_Forge/stable-diffusion-webui-forge/venv/lib/site-packages/bitsandbytes/libbitsandbytes_cuda128.dll') WARNING:bitsandbytes.cextension:The installed version of bitsandbytes was compiled without GPU support. 8-bit optimizers, 8-bit multiplication, and GPU quantization are unavailable." "No module named 'triton.ops'"

Hockey

Which file do you input these commands?

SECourses

please read 8 August 2025 Update

SECourses

please read 8 August 2025 Update

Kevizz Men Of The World Art Adult/18+

I have Forge version: f2.0.1v1.10.1-previous-669-gdfdcbab6  •  python: 3.10.6  •  torch: 2.3.1+cu121  •  xformers: N/A  •  gradio: 4.40.0  •  checkpoint: 57bde3fc3a. Will this just update it?

JustDrwIt

what a goddamn joke this was. wasted 2 hours of my friday night

SECourses

delete venv and make a fresh install

Roni Bliss

I think you should just give up mate, Clearly sounds like following a simple tutorial where every step is clearly explained to you is just too much for you to handle! A whole 2 hours? ..... On a Friday night.? Faruk ... Look what you have done, You broke his brain! Hope you're happy with yourself!

Fernando Martin

Hi Furkan. Is there a way to bring back the old Flux Realistic samplers? i ran the command git reset --hard f53307881bfd824dbdce6ac0d4bba04d9a74ab36 in the webui folder, but it seems to break the install. After that i get this error: RuntimeError: Failed to import diffusers.pipelines.pipeline_utils because of the following error (look up to see its traceback): Failed to import diffusers.models.autoencoders.autoencoder_kl because of the following error (look up to see its traceback): No module named 'triton.ops' Press any key to continue . . . Thanks. Great work

SECourses

then you need to debug and find out working versions it is a lot of work. i can try to fix in a consultation session but currently really busy atm for this

Hockey

Unfortunately it seems like we have to choose between a fast Forge with the samplers or slower Forge with samplers. I have the exact same error as you Fernando. But maybe the other samplers can compete with the Forge samplers with Flux Krea? Thank you Furkan for updating, really appreciate it and have a wonderful vacation. It's okay and understandable for you to be absence for vacation. ;)

SECourses

i dont know i am using iPNDM and it is the best sampler for FLUX and you are welcome

Fernando Martin

i was trying yersterday and found out that a recently previous version that Furkan uploaded, Forge_installer_v9 and it worked. This is, faster Forge with Flux samplers (you need to apply that git command on the webui folder and start the app from the webui-user.bat) but then i installed the v12 that is a lot faster, but when i try to bring back the Flux samplers this latest version gives me this error. I guess i will go back to v9.

SECourses

well our latest zip file installs the latest repo with torch 2.7. perhaps torch 2.7 not supporting what you want not sure

Fernando Martin

Furkan, not trying to say you are wrong but, i have a working version of Forge with the Flux samplers and believe, nothing beat those. It's a shame they left those out

SECourses

i tested all samplers on swarmui so probably swarmui doesnt have it in that case.

Fernando Martin

@Hockey after a few hours i got it. -Install the latest version (v12, this one). -go to your webui folder. in the address bar of your explorer type CMD, to open a console in that location. -copy and apply this command: git reset --hard f53307881bfd824dbdce6ac0d4bba04d9a74ab36 After that, you will get an error. So delete the VENV folder and install again Forge from the .bat that Furkan uploaded. (install it in the same location). Now start Forge from the webui-user.bat and there you have. Blazing fast Forge WITH Flux Samplers. NEVER UPDATE AGAIN :)

SECourses

thanks i will test and compare this :D which flux sampler do like and see as best?

Hockey

I actually solved it thanks to a thread here: https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/2554 I just came to tell you. =D Anyway to anyone that need the Forge Flux Realistic samplers back in v12 this is an alternative to Fernandos solution: Go to: https://github.com/lllyasviel/stable-diffusion-webui-forge/commit/b835f24a4d9732c936a3d41d62adfae46bd82441#diff-b8454e8123e62f9ec2c0d4e299e0690c27592af4230a16976ecd4745e61871e7 Re-add the deleted (red) lines into initialize.py and launch_utils.py which is located in: stable-diffusion-webui-forge\models Then go to folder: stable-diffusion-webui-forge\modules_forge In notepad paste this: # See also: https://github.com/lllyasviel/google_blockly_prototypes/blob/main/LICENSE_pyz import os import gzip import importlib.util pyz_dir = os.path.abspath(os.path.realpath(os.path.join(__file__, '../../repositories/google_blockly_prototypes/forge'))) module_suffix = ".pyz" def initialization(): print('Loading additional modules ... ', end='') for filename in os.listdir(pyz_dir): if not filename.endswith(module_suffix): continue module_name = filename[:-len(module_suffix)] module_package_name = __package__ + '.' + module_name dynamic_module = importlib.util.module_from_spec(importlib.util.spec_from_loader(module_package_name, loader=None)) dynamic_module.__dict__['__file__'] = os.path.join(pyz_dir, module_name + '.py') dynamic_module.__dict__['__package__'] = module_package_name google_blockly_context = gzip.open(os.path.join(pyz_dir, filename), 'rb').read().decode('utf-8') exec(google_blockly_context, dynamic_module.__dict__) globals()[module_name] = dynamic_module print('done.') return Save as: google_blockly.py Then you should be able to start as usual and the Forge Flux samplers should be visible.

Fernando Martin

@Furkan, all 4 Flux Samplers are very interesting. Most than all "legal" samplers. Even the nice IPNDM (i was using DEIS). In my testing, bests results are with Flux 2x + DDIM (Those 2 require more steps than the others)

RC123

Hey, do I need to download the latest file (Forge Installer V12.zip) to get the update or can I just click "Windows_Update_Forge_Web_UI.bat" in my existing folder?

SECourses

please make a fresh install for this case. since we made some serious changes

Bryan Henson

Just did a fresh install on this. I'm using a 5060 Ti 16gb. I got this error on my first generation trying to use an SDXL model. 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.

Bryan Henson

That's actually what I did; I followed your entire video. When i ran your install script, it said it was using 3.10 but I will check again. Anything else I can look at if 3.10 was used?

SECourses

it is best to send me install logs. monstermmorpg@gmail.com please make a fresh instlal so i can see all

Fernando Martin

Hi @Hockey did you see the newest version of Forge? it seems we'll have to find a way to add Flux Samplers again :*) Neither of our resources work in this one. The folder structure is different and the initialize and launch files are also different. :/ Let me know if you find something. I will let you too

Pew

C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12>set HUGGING_FACE_HUB_TOKEN=hf_****************************** C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12>cd stable-diffusion-webui-forge C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge>set COMMANDLINE_ARGS=--skip-requirements C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge>REM put command line args next to COMMANDLINE_ARGS= e.g. set COMMANDLINE_ARGS=--skip-requirements --always-normal-vram --disable-xformers C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge>REM --skip-requirements flag will skip reinstalling requirements_versions.txt each time C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge>webui-user.bat venv "C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge\venv\Scripts\Python.exe" Traceback (most recent call last): File "C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge\launch.py", line 4, in from modules import launch_utils File "C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge\modules\launch_utils.py", line 17, in from modules import cmd_args, errors File "C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge\modules\cmd_args.py", line 6, in from backend.args import parser ModuleNotFoundError: No module named 'backend' Press any key to continue . . . C:\Users\Nova\orbital\ml-lab\Forge_Installer_v12\stable-diffusion-webui-forge>

SECourses

did you follow this tutorial first? https://youtu.be/DrhUHnYfwC0 i also need full logs for Windows_Install_or_Update_Forge_Web_UI.bat in a fresh new folder. email me : monstermmorpg@gmail.com

SECourses

ok i made a fresh install. the installation was broken. fixed it. now it works with a fresh install same zip file