This is a model trained on caves, and caverns. I made this because I thought the default caves looked horrible or nonexistent. May need to increase it starting at 0.6 for desired results. This model was trained on NAI anime-full-pruned and generated in anything5. Made with the Kohya tool for AI Training through google colab. Any suggestions to improve or recommended things for the data set would be appreciated.
Description
Trained on 55 images of real and anime caves, with a mixture of underground water sources as well.
FAQ
Comments (4)
I'd be curious to see if I could retrain with your dataset but instead of dreambooth, I would finetune a checkpoint. If you're interested, let me know.
Oh do you need the training data? I just added it to the downloadable files.
@rulles I would have re-run your dataset through finetuning and send you the checkpoint. If it's only the captions, I have little use for it as I already have a lot on my plate.
@rulles Nevermind, I just noticed that you uploaded your dataset, I'll run it for finetuning and post a link of the checkpoint unless you prefer that I send it to you directly. My reddit username is the same as this username so you can send me a message there and I share privately share the link to the trained checkpoint.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.