how many gigs of vram does pixart use? everytime i run app.py, it says out of vram
how many gigs of vram does pixart use? everytime i run app.py, it says out of vram
MI just trained with a very interesting set of images. Even though I selected the best photos, I could only find the best d_coef value for a quarter of them. I tried 0.85, it didn't come out strong. With 1.02 it was too strong, and even with 0.95. Finally, with 0.9 I managed to get a near perfect Lora. On another set of images, I managed to get one that turned out perfect the first time, even though there was nothing special about the photos. So there are still some exceptions where, even if the graph is good, you still need to polish the final result a bit.
Fi am following their repo
Fthey havent released yet
Fthat is very best PC
Fsurely enough
Fjust ignore them. tensorRT spits them out but works
Fyou set default cache folder
Fand they get downloaded into there
Fhugging face cache folder
Flook at the runpod instructions it shows
Fhow much vram you have?
Z12
Fcan you try low vram file
Fi added
SPerfect, thank you! Love your work
ZI will thank you
Fif you are on linux just look at the hyper parameter when starting the app.py
Flet me know result please
ZWill do
W@Dr. Furkan Gözükara i am on the 2nd episode of your course on udemy, have you in latter courses also specified how to prompt stable diffusion to generate amazing images like how people do (chubby:0.4) something like that
DI am training my photos for sdxl lora today. I used ohwx woman for my wife , can i use same ohwx man ? or ohwx should be different ?
Dok i am using bbuk man for me and ohwx woman for my wife
Dthis should help me to create photos of us nboth i think
DNeed clarity on sdxl Lora training kaggle notebook code
DThere is different run code for Lora training after koyass webui run
DIt's giving me old error
RHi, @Dr. Furkan Gözükara some questions about Kohya training for SD. I would like to train clothing based on a culture. Basically, using any checkpoint, I can just invoke them to wear the clothing with just a keyword. Would a Lora be the best option? Also, when training, how do you best caption the dataset so that the Lora does not remember the background, the face, color of the clothing, physical weight of the person, etc. I only want the clothing, and want to be able to determine the colors. How many images would be minimum for clothing? And what's the best way to caption? I have tried so many different ways, and haven't been very successful.
Bhttps://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.7.0-RC
Automatic1111 is soon releasing 1.7.0
Automatic1111 is soon releasing 1.7.0
GitHub
How to use: wiki.
Features:
settings tab rework: add search field, add categories, split UI settings page into many
add altdiffusion-m18 support (#13364)
support inference with LyCORIS GLora netwo...
Features:
settings tab rework: add search field, add categories, split UI settings page into many
add altdiffusion-m18 support (#13364)
support inference with LyCORIS GLora netwo...
Dok i think i found the issue (possibly)
In json , there is add parameter ""additional_parameters": "--max_grad_norm=0.0 --no_half_vae --train_text_encoder","
removing --train_text_encoder solved teh sdxl_train.py error
i might be wrong , or this text shouldnt be removed.. @Dr. Furkan Gözükara should only confirm
In json , there is add parameter ""additional_parameters": "--max_grad_norm=0.0 --no_half_vae --train_text_encoder","
removing --train_text_encoder solved teh sdxl_train.py error
i might be wrong , or this text shouldnt be removed.. @Dr. Furkan Gözükara should only confirm
Dsince we are doing lora training and this parameter seems to be for dreambooth , this is causing the error likely
DKindly provide two different notebook files , one for dreambooth and another for lora , so that it wont confused , also the dreambooth_best json file keep overwriting the locally saved , whenever we need to rerun the kohyaass gui. Wish there was other option
Dso when i use two different face trained lora , how am i supposed to use adetailer ? and that too correctly on male and female faces ?
Fi think i had shown it but cant remember in which lecture it was
Fyou should have seperate model
Dlike ?
Fif you can collect 1000s of images
Fyou can caption with llava
Fand really obtain improved model
Fi would do dreambooth and then extract lora instead of lora training
Fi am reading changes
Fuse latest configs
Fthey are made for latest gui
Dmy wife lora is ready.. "
I am creating mine now.
So lets say i use something like
<lora:wife:1>,(ohwx woman) standing with <lora:husband:1>,(bbuk man) in an island
I am creating mine now.
So lets say i use something like
<lora:wife:1>,(ohwx woman) standing with <lora:husband:1>,(bbuk man) in an island
Dit will create two faces , in adetailer what prompt should we give ?
Di think it was there in notebook
Dit was inherited from "Kaggle_SDXL_DreamBooth_Best.json"
Dif we want to train lora , we cant skip moving this file to our folder if we run the code exactly
D
DSomeone help me with creating images with two different face lora