Furkan Gözükara
patreon
PixArt-alpha (PixArt-α) Automatic Installer For Both Windows And RunPod With Additional Web UI Features
🕑 Added 2023-11-27 14:06:22 +0000 UTCJoin discord and tell me your discord username to get a special rank : SECourses Discord
RunPod Register / Login Link , RunPod Beginners Tutorial
16 January 2024 Update
8-bit loading error fixed
Library versions updated
3 December 2023 Update
Please redownload the zip file
Now you will see lowVRAM_1024. bat file
This will off load some parts into CPU to further reduce VRAM usage
I was able to generate images on RTX 3060 with a little bit of shared VRAM usage
If you have already installed the repo run update. bat file first to get latest version of app. py from our repo
Tutorial video released : https://youtu.be/ZiUXf_idIR4?si=c966ckIS8TjEZlNg
PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis
PixArt-α is the true new rival of Stable Diffusion XL (SDXL)
The quality is amazing and it is also fast
Tutorial Video : https://youtu.be/ZiUXf_idIR4?si=c966ckIS8TjEZlNg
Please also upvote this Reddit thread if possible
Requirements:
Make sure that you have git and Python 3.10.x installed. I used Python 3.10.11
Here a tutorial video : https://youtu.be/B5U7LJOvH6g
If you encounter any network related problems during install or model download use https://1.1.1.1/ WARP VPN of Cloudflare which I am using and totally free.
How To Install And Use
Download pixArt_installer_v2.zip and extract into any folder where you want to install
It will generate a new venv so it will not affect your other apps such as Automatic1111 SD Web UI
Run install.bat on Windows and after that use 1024_16bit. bat or other bat files
For RunPod please read runpod_instructions_READ. txt file
The running will download models into the Hugging Face cache folder
Features of Improved Gradio Web APP
The improved Web APP Supports automatic saving of generated images into outputs folder
A new parameter batch count which lets you to generate any arbitrary number of images with same prompt like set it 100 and it will generate 100 images
Supports loading in 8bit, 512 pixel model and Dall E3 VAE
If you want Hugging Face to download models into another drive (not mandatory) use following prompt
Start a new CMD as administrator
setx HF_HOME "G:\HF_Models"
This above command will add a variable to System Environment Variables
Comments
Furkan Gözükara
your gpu is not NVIDIA CUDA that must be the reason. please use runpod. i will hopefully make massed compute installer which is much better than runpod
Eugene Mazarakis
Aftell the installation, i was tried to create images and this was possible. I ran the 1024_16bit
Eugene Mazarakis
I used install.bat. My python version is 3.10.6. Mu GPU is Intel(R) Iris(R) Xe Graphics. I cannot create images through the local url. How can i find the "per step image generation speed"?
Furkan Gözükara
hello i just did a fresh install and works perfect. what is your gpu? what is your per step image generation speed? what is your python version? i used install.bat file
Furkan Gözükara
no that means install error. i am testing now. i will also update this for pixart beta actually
Eugene Mazarakis
The 1024_16bit runs on a cpu? The gradio web-ui returns a message "Running on CPU �� This demo does not work on CPU."
Furkan Gözükara
Hi i should test it whether possible or not. not sure at the moment. stay tuned
Pallavi Chauhan
Hi, Can we run PIXART-α on kaggle notebook?
Furkan Gözükara
I deleted very big message. but you have error D:\ComfyUI_windows_portable\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\python_embeded\python.exe: No module named venv The system cannot find the path specified. now i need you to show me 2 things open a cmd window and type python and show me screenshot of in which folder you are installing pixart
Jonathan Bach
same error when attempting to run The system cannot find the path specified. Traceback (most recent call last): File "G:\PixArt\pixArt installer\PixArt-alpha\app.py", line 6, in import gradio as gr ModuleNotFoundError: No module named 'gradio' all started Press any key to continue . . .
Furkan Gözükara
install error. while installing what did you see on cmd? so at some point it stopped and didnt install rest. if you record installation video or install cmd screenshots i can tell you error
Jonathan Bach
getting this error when run the bat file The system cannot find the path specified. Traceback (most recent call last): File "G:\PixArt\PixArt-alpha\app.py", line 6, in import gradio as gr ModuleNotFoundError: No module named 'gradio' all started Press any key to continue . . .
Furkan Gözükara
nice congrats
gao yu
finally sucess in docker
gao yu
ok, i will try to delete all the cache
Furkan Gözükara
i see. i think that is related to gradio. you may ask on their github : https://github.com/gradio-app/gradio
quirkformity
Just to make it available on my local network. Automatic1111 and others report "0.0.0.0/7860" as the address when the "--listen" argument is passed.
Furkan Gözükara
why do you want to get it binded to 0.0.0.0?
quirkformity
Yes. It works fine as a localhost address, but the BAT script's arguments aren't getting passed to the app. That is, it binds to 127.0.0.1, but I can't get it to bind to 0.0.0.0.
Furkan Gözükara
It is already running locally on Windows?
quirkformity
Apologies. Local (Windows).
Furkan Gözükara
You mean using locally on runpod?
quirkformity
Adding "--listen" to COMMANDLINE_ARGS doesn't work. Is there another method for making it available on a local network?
Furkan Gözükara
that is weird. comes to my mind is you had error while downloading the model files into cache. can you delete them and redowload?
Furkan Gözükara
sadly MacOs will not work. you can run it on RunPod. you know Mac is extremely limited for AI and Machine learning. even AMD gpus will not work
Carsten Li
Could you please also add a MacOS script?
gao yu
i upgrade to 3.10.11 , but still unsucessfull, " return cls._from_pretrained( File "E:\pixArt installer\PixArt-alpha\venv\lib\site-packages\transformers\tokenization_utils_base.py", line 2073, in _from_pretrained raise ValueError( ValueError: Non-consecutive added token '' found. Should have index 32100 but has index 0 in saved vocabulary."
gao yu
i follow your tutorial video , may be python 3.10.6 not work
Furkan Gözükara
ok now that tutorial released can you follow it and let me know?
Furkan Gözükara
thank you so much
AI Made Me Rich 🤑
🤩 wow phenomenal thank you 🙏🏽
gao yu
my python version is 3.10.6, all install completed,and no error
Furkan Gözükara
i hope too it is amazing model
mypatreonemailacc
I hope automatic1111 supports it.
Furkan Gözükara
that is my next thing to look hopefully
mypatreonemailacc
Are there scripts to train a Lora or TI for PixArt?
Furkan Gözükara
yep so many new stuff
Samuel
Perfect... There are a lot of news currently regarding AI. Im a bit behind lol
Furkan Gözükara
not yet. they say they have training script too even dreambooth
Furkan Gözükara
what is your python version? also were models accurately downloaded are you sure of that?
Samuel
Wondering, did you try training on this?
gao yu
i use 1024_16bit. bat after sucessfully run install.bat
gao yu
how to solve this:ValueError: Non-consecutive added token '' found. Should have index 32100 but has index 0 in saved vocabulary.
Furkan Gözükara
i don't know yet. but they say supporting controlnet too
Tiago4D
Hi! Could I use it with img2img or inpainting?
Furkan Gözükara
sure that would be great. i plan to record video hopefully today
Rasika Singal
I will add your video to my article, that promotes your work. Please let me know if that is ok?
Furkan Gözükara
nice one
Rasika Singal
Please check my article on it, https://medium.com/illuminations-mirror/exploring-pixart-α-fast-training-of-diffusion-transformer-for-photorealistic-text-to-image-2226fc677152
