Blaziken LoRA trained on NovelAI with Kohya training scripts using the DAdaptAdam optimizer at 512x512 resolution. Has trouble getting the feet right because they're not present in most images in the training set. Tail placement and shape can be unstable. Should have wide compatibility with model mixes based on NovelAI, which is most of them. Trained on an all female dataset.
Description
Initial release
FAQ
Comments (6)
Hey so I was testing this out on my own computer and was comparing it to other Loras and I can't help but notice that when I used Tomas_Aguilar's Blaziken Lora the results were exactly identical with no prompt change and using the same steps and seed and cfg, even on different checkpoints. Any idea what's going on there?
If you're getting the same result no matter what LoRA you're loading, then it's likely there's something in your setup preventing LoRAs from loading in at all, or some other error in your software setup. An easy way to troubleshoot this would be to load a non-blaziken LoRA and see if you still get the same output. Our models use completely different training data and training settings, resulting in two different models with varying filesizes (18 MB for Tomas' model, ~144 MB for mine). My model is certainly not a reupload of someone else's work, if that's what you're insinuating....
@KomodoStar I wasn't trying to insinuate anything I was mostly just confused because I've never seen that happen before. I did a full reset and it appears something must've gotten stuck somewhere because things are now working as expected with getting completely different results. Sorry for causing any issues.
@Bluefurvert No worries, it’s all good! I’m glad you were able to get things fixed. I hope you’re enjoying the model!
@KomodoStar Oh I absolutely have, thank you so much for it