Thats me :)
You can find all of the models on my hugging face. Might i ask where you got those if not from my hugging face im curious?
Oh hello, this comes from your Hugging face? I just download models people to tell me to, I don't remember seeing it it from hugging face, maybe a google drive?, anyways I love your models I've used them and they are nice,
just saw your models from you hugging face. do you want me to remove them? and just link it to your hugging face?
@scriche
also if you made these on your own may I know how :D I've always wanted to make my own models and stuff
🤔
Naa its fine you can leave this up i make them to be used and shared by everyone :). I use Automatic1111 Webui with the dream booth extension and usually have around 300 pictures and set on the dream booth concepts tab the dataset directory of all the images instance prompt of [filewords] which grabs all the tags from the .txt file associated with the image which are acquired by using my own python scripts to scrap the tag+image of danbooru though you can use a new program call (Danbooru downloader) which is probably easier never tried it. i dont set any class image stuff. for general setting i set the raining steps per image (Epochs) to 200 which will train each image 200 times which would be 200 * 300 images so 60,000 steps which will take about 6 ish hours on a 3080 ti 12GB VRAM which you will need a card with a decent amount of VRAM. everything else i keep default and in the end i usually merg the RAW model back into anythingV3 at 15% anythingv3 and 85% new model to get cleaner faces and a sharper image. Hope you make some cool models
@scriche ill link it to you just in case they would to thank you :D
I use scraper too, but haven really created a model :) maybe I can ask you later when I would like help :)
Sure sounds good as im always uploading and updating the models. Feel free to contact me if you got any questions :)
@scriche love you :D