LoRA from Gundam Hathaway by Gigi AndaluciaUpdate logV1.0First versionV1.0 halfSince there were comments stating that the initial version was overfitted, I am releasing a version trained for half the number of steps—an intermediate checkpoint generated during the original training process. I have no plans to recreate the model unless I personally feel the need for it.V1.1Train on a new dataset of 100 imagesPlease do not make any requests.Creating a LoRA model is a process that takes at least one day, from building the dataset to adjusting the labels, training the model, and verifying its performance.<hr />I often receive comments like these after releasing a model:"Please make a character named [Name].""Do you plan to make a [Name] version?"I find receiving comments like this incredibly tiresome.As stated in my profile, please do not make model requests. Do not ask for new characters or new versions. I am simply sharing the results of a project I pursue for my own enjoyment.Asking "Do you have plans for a new version?" when I haven't announced anything is essentially demanding a remake. It is extremely presumptuous.Please don't act as if you are critiquing a meal after someone has kindly shared their packed lunch with you; this isn't a restaurant.If you aren't satisfied with it, please find something that suits you or create it yourself.<hr />

Gigi Andalucia

Training with 300 images network 32/16 unet only,lr : 0.0001

Files with SHA256: edb7c83d2c4207e76bf6d16b56174e00a0e58750d01541e97eadf819a3681214