Furkan Gözükara
patreon
Forge Web UI Latest Version RunPod Auto Installer and FLUX Auto Model Downloader for Windows, RunPod and Massed Compute
🕑 Added 2024-08-18 12:28:23 +0000 UTCthis post is not maintained anymore here newer version : https://www.patreon.com/posts/118442039
this post is not maintained anymore here newer version : https://www.patreon.com/posts/118442039
this post is not maintained anymore here newer version : https://www.patreon.com/posts/118442039
this post is not maintained anymore here newer version : https://www.patreon.com/posts/118442039
this post is not maintained anymore here newer version : https://www.patreon.com/posts/118442039
this post is not maintained anymore here newer version : https://www.patreon.com/posts/118442039
this post is not maintained anymore here newer version : https://www.patreon.com/posts/118442039
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)
How to Install and Use on RunPod
Just download the attached zip file Forge_Installer_v2.zip and extract into any folder on your computer or on your phone.
Open Runpod_Instructions_READ.txt file
According to the file select accurate Pytorch 11.8 official template with any GPU, set your volume disk size.
After started upload all files into work space and copy paste the commands into a new terminal. It will auto install latest version of Forge Web UI and auto start it.
Just read the Runpod_Instructions_READ.txt file carefully for all info.
To automatically download FLUX models read Runpod_Instructions_READ_FLUX_Download.txt and follow the commands. It will ask you to download which model like below.
For 24 GB GPUs like RTX 4090 I suggest Download FLUX.1-dev FP8 Version - best quality
Proper launch arguments auto set for 24 GB GPUs on RunPod like --cuda-malloc --always-gpu
RTX 4090 FLUX 1024x1024 FP8 dev model speed is 1.6 it / second and immediately starts generating next image - no unload or reload
More FLUX auto download options will be added later hopefully
Don't forget to download T5 Text Encoder. It will also download Clip L. VAE automatically downloaded when you download FLUX model
Installing requirements takes time so patiently wait
How to Download FLUX Models
For Windows, put files into the parent folder of stable-diffusion-webui-forge and double click and run Windows_Start_Download.bat file
For Massed Compute put anywhere it will auto download into accurate folder in our SECourses image
For RunPod, upload all files into workspace and it will work
Comments
Furkan Gözükara
this installs latest version always but let me test
Yannis
Hey mate, could you offer an updated version ? :)
Furkan Gözükara
nope you dont need. it should work
Daniel Cardona Ramirez
Do I need to do anything to use Lora created in Kohya in Forge?
Furkan Gözükara
no they are just for reading them if you want to learn more
C
Hi I'm wondering what this part means in the Runpod_Instructions.txt mean: Forge command line args can be found below check both links : https://github.com/lllyasviel/stable-diffusion-webui-forge/blob/da72cd6fabd149fd0752c63e5a84551869308c6d/backend/args.py#L55 https://github.com/lllyasviel/stable-diffusion-webui-forge/blob/da72cd6fabd149fd0752c63e5a84551869308c6d/modules/cmd_args.py#L47 Do I have to paste those links in somewhere, in a specific way or what do I do with that ?
Furkan Gözükara
yes please do both and it should work.
piXtar playz
I had the same error. Do I have to download the zip again to get also the correct setting? Thx for your work btw. :)
Rafael
Great, thank you!
Furkan Gözükara
i auto edit it for you in the installer. that is why it works fast
Rafael
Thank you for the screenshot with specifications! I don't know what I did wrong (I'm just an amateur) but strangely, when I tried everything again on a new pod (also 48 GB) I still got the same message BUT now it did generate the images (at about 35-36 GB usage). But I must admit I didn't know what you meant in the tutorial file when you said "You need to edit webui-user.sh file after above install script run before executing run command below for adding removing args", so I ignored that - maybe that causes the issue?
Furkan Gözükara
I just tested with a 48 GB GPU and it used maximum 32.2 GB VRAM. how did you get error so weird. here my config https://pasteboard.co/e6WMUMfuJvZl.png
Furkan Gözükara
Probably they changed some config let me test and fix
Rafael
Hi, at the very end of my installation in Runpod, just after I have gotten my live Gradio link, I get this message (in my terminal): "You just set Forge to use 100% GPU memory (47645.00 MB) to load model weights. This means you will have 0% GPU memory (0.00 MB) to do matrix computation. Computations may fallback to CPU or go Out of Memory. In many cases, image generation will be 10x slower. Make sure you know what you are testing". And indeed, when I try to generate an image (even just 200 x 200), I get an out of memory warning after just 2 seconds.
Furkan Gözükara
just did fresh install on rtx 4090 works perfect 0 issues https://ibb.co/d749X0Z https://ibb.co/r61x50C
Furkan Gözükara
which template did you select exactly? :2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04 i am going to test again now to be sure
John Robert Planta
I get this error when trying to run in Runpod: I followed all instructions. Traceback (most recent call last): File "/workspace/stable-diffusion-webui-forge/launch.py", line 54, in main() File "/workspace/stable-diffusion-webui-forge/launch.py", line 50, in main start() File "/workspace/stable-diffusion-webui-forge/modules/launch_utils.py", line 546, in start import webui File "/workspace/stable-diffusion-webui-forge/webui.py", line 19, in initialize.imports() File "/workspace/stable-diffusion-webui-forge/modules/initialize.py", line 32, in imports from modules import processing, gradio_extensions, ui # noqa: F401 File "/workspace/stable-diffusion-webui-forge/modules/ui.py", line 16, in from modules import sd_hijack, sd_models, script_callbacks, ui_extensions, deepbooru, extra_networks, ui_common, ui_postprocessing, progress, ui_loadsave, shared_items, ui_settings, timer, sysinfo, ui_checkpoint_merger, scripts, sd_samplers, processing, ui_extra_networks, ui_toprow, launch_utils File "/workspace/stable-diffusion-webui-forge/modules/ui_settings.py", line 10, in from modules_forge import main_entry File "/workspace/stable-diffusion-webui-forge/modules_forge/main_entry.py", line 29, in 'float8-e4m3fn': (torch.float8_e4m3fn, False), AttributeError: module 'torch' has no attribute 'float8_e4m3fn'
Furkan Gözükara
no in seperate repos. you can see all links of them here : https://www.patreon.com/posts/110293257 Models Links Downloads
Steve
Finally got this to work with my 4070 except it doesn't show clip_l.safetensors and t5xxl_encoly.safetensors are they in the original repo?
Furkan Gözükara
ah ye forge ui is in development mode old things are not working. he is moving into flux era
V.P.M Video Projection Mapping
I have asked in that discord channel and they tell me it is the wrong place. I have this problem with Forge, not with SwarmUI.
Furkan Gözükara
please ask this in swarmui dev channel they answer very quickly : https://discord.gg/swarmui-1243166023859961988
Furkan Gözükara
hopefully you will be able to i am in research.
Jason Dalton
If I only have 16GB of VRAM can I still do this? I have 32GB of Total RAM and have a 4070 Ti Super
V.P.M Video Projection Mapping
I've downloaded the Tileddiffusion extension to scale images. When I click "Apply and restart UI" the browser gets stuck with the message "Reloading..." Is this normal? Has this happened to anyone else?
Furkan Gözükara
wow interesting nice
Furkan Gözükara
ye fp16 will do offloading :d or you need to get bigger vram
Sipriyani
this was the most important notice, i should have read : * RTX 4090 FLUX 1024x1024 FP8 dev model speed is 1.6 it / second and immediately starts generating next image - no unload or reload so fp16 load/unload takes more than 30 secs for me which stumbles my patience. :) downloading fp8 now..
Yannis
I solved the problem with a different GPU. Switched from a RTX4090 to a RTX 6000 Ada. Loras are now working. Also I found the loras to have an effect on the image only when picking dev-fp8 as checkpoint (with nf4 impact is very limited to non existent)
Furkan Gözükara
great
Furkan Gözükara
which base model you using? quant models not working with loras. i dont know if forge added lora support for any model
Furkan Gözükara
which base model you using? quant models not working with loras. i dont know if forge added lora support for any model
Yannis
Are Loras working for you guys ? None are working for me
Sipriyani
works like a charm. thanks
Sipriyani
Yes it was 7z file that I have downloaded and only extracted.
Furkan Gözükara
thank you updated to v2. can you verify it is working? thank you so much
Furkan Gözükara
hello. did you use pre installed package? i am downloading to verify path will fix ty
Sipriyani
I needed to edit windows bat file for folder typo "if not exist "stable-diffusion-webui-forge\modules_forge" ....." Instead.. stable-diffusion-webui-forge\webui\modules_forge\
