i will share amazing script as well that can auto crop subject into any aspect ratio you wish
i will share amazing script as well that can auto crop subject into any aspect ratio you wish
AHi everyone! Trying to do dreambooth training, but keep getting this error. Could anyone help me with this one?
message.txt9.89KB
Ftry another optimizer. looks like it did run like 300 steps
Falso dont save samples
Flooks like broken atm
Athanks, i'll try!
Nany luck in getting it to work?
Fi did a training yesterday with new main branch
Fthey merged
Fon windows
Fwill test runpod soon
FTraining set posted on patreon
FI will make a video about it
QAutomatically or manually?
AIt works fine
AWhere is this option located?
FFully automatically
Qthats great
UI can not see clip vision preprocessor on the list.Does not show up.Also hut is apsent.Could you please tell me hoe can I solve this problems.Thank in advance.
Fah they changed some names
Fdid you update extension to latest version
Fi had been getting some blue screens. i think it was related to dram voltage. raised it to 1.39 
UYes Everything is up to date.By the way Are you Turkish?
UI just found your tutoril "Easiest Way to Install & Run Stable Diffusion Web UI on PC by Using Open Source Automatic Installer".I think I gonna install everything new according to your tutorial.
GGThe "Dreambooth" extension seems to be gone
GFyes I am Turkish
Fyes git clone and install manually
Etrying Lora training with Kohya, im getting "AttributeError: module 'tensorflow' has no attribute 'io'" error. any ideas?
Fwhat python version you have?
E3.10.9
Ei watched ur python / venv video, im still having trouble grasping it, need to watch it a couple more times lol
Flooking good. i wonder if cuda toolkit is too high version
Fi would try reinstall from scratch
Fbut before
Ftry different optimizer
Eohh yeah mayube it it actually. okay il ltry reinstall now.
Fwhich one you tried?
Ftry adamw
Fnot adamw8bit
Q@Dr. Furkan Gözükara I have accumulated 500k images for faces and 100k images for landscape photography.
I think we can achieve realistic generations with such a huge dataset.
I want to divide each dataset into multiple mini-datasets so that we can train mini-models of each and later merge the mini-models into a single huge model. I am looking for collaborators because it is going to be a huge model, we need a team to make it efficient and quick. I am even and for suggestions. I am looking to caption the images using minigpt4.
Can you please collaborate or is anyone interested?
I think we can achieve realistic generations with such a huge dataset.
I want to divide each dataset into multiple mini-datasets so that we can train mini-models of each and later merge the mini-models into a single huge model. I am looking for collaborators because it is going to be a huge model, we need a team to make it efficient and quick. I am even and for suggestions. I am looking to caption the images using minigpt4.
Can you please collaborate or is anyone interested?
Fwell i dont know anyone but you can post this time to time and see if anyone is interested in
QYeah, someone can look upto it
Qwant to collaborate with people who can help me with training as I am new, to training I am looking for people who have a gpu so that we can use the machine power to train the model as it cannot be trained using 1 or 2 gpus(if thats the case it would take forever) and we even need to caption them so we gpu power for that.
the dataset is gonna be split into multiple datasets, so those mini datasets can either be trained on colab or locally through gpus.
I want to make faces realistic so that SD can come close to MJ. In SD as of now, all I see is waifu faces. Waifus faces are polished and dont have much detail in their faces like freckles, spots, etc. In case of male faces they dont have facial hair.
Yep I am planning to include NSFW as thats the superior thing that SD has that no other image generators have
I had accumulated 700k images from various subreddits.
the dataset is gonna be split into multiple datasets, so those mini datasets can either be trained on colab or locally through gpus.
I want to make faces realistic so that SD can come close to MJ. In SD as of now, all I see is waifu faces. Waifus faces are polished and dont have much detail in their faces like freckles, spots, etc. In case of male faces they dont have facial hair.
Yep I am planning to include NSFW as thats the superior thing that SD has that no other image generators have
I had accumulated 700k images from various subreddits.
E@Dr. Furkan Gözükara no matter what i do i can get dreambooth extension to successfully create Lora and the Kohya gui wont work for me either. you do remote assistance/installation if were a part of your patreon ?
Fbecome a silver patreon please i will connect and install
Eokay. I am doing so now.
Fty dming you
DI’ve seen various workflows on Reddit but either they don’t work or are hard to follow and understand
