5KilosOfCheese
5KilosOfCheese
Beam search branching method of searching, where at every layer you have all the possibilities (in this case tokens). Instead of choosing just 1 option which has the highest score,...
Can your provide the following things: 1. The training settings you used (preferably the "print training command" output, by just copy pasting it. 2. General overview of the samples, the...
> I am having the same issue, it sometimes outputs good images but mostly the images are completely garbled. > > here are the training settings I used `kohya_ss/sd-scripts/sdxl_train_network.py" --bucket_no_upscale...
See if this helps https://github.com/bmaltais/kohya_ss/discussions/1744 This is a thing which happens to many. The safe bet is that you remove ALL symbols and letters that aren't the standard English set....
You seem to have tried to do DreamBooth training. In the GUI select the LoRA tab and see if that works.
You uncheck the use of WANDB in the advanced configuration:  This will then default to Tensorboard. You can then disable logging alltogether, by leaving the Log director path empty...
> 5KilosOfCheese many thanks for your answer but I'm confused: I'm training an SDXL lora, my WANDB in advanced configurations is unchecked and the Log director path is empty but,...
> One more question while I have the attention of the masters: about the presets in KOHYA SS (prodigy, LOHA, LOKR, kudou reira...) is there a place where I can...
> Yea, maybe criteria for displaying them is for a little presentation of what the settings do, and some insight on the main differences to basic settings. Would be useful...
> @5KilosOfCheese I could certainly add this as documentation in the repo for users to access in the GUI if you create it… you would need to learn the basic...