The training dataset consists of screenshots from her re-debut (2.0 model) and 3d showcases. Given the nature of the dataset, you might have to do some further upscaling when generating full body shots. The lora works best for upper-body shots, at least on weights close to 1. On rare occasions it will generate the wrong hair color, just add it as a negative prompt if it ever happens. Some details about training: <ul><li>dataset_count = 66</li><li>optimizer = CAME</li><li>learning_rate = 0.0004</li><li>learning_rate_scheduler = cosine</li><li>epochs = 11</li><li>steps = 726</li><li>lora_rank = 32</li><li>lora_alpha = 1</li><li>batch_size = 1</li></ul>

IRys - Hololive (Illustrious v1.0)

This lora features my favorite IRys outfit, and my favorite combinations of it. It has only 2 hairstyles (bob cut and ponytail), the beret hat, glasses, and her off shoulder dress. The cardigan and twintails hairstyle were not part of the image dataset.The training dataset consists of screenshots from her casualRys outfit reveal. Given the nature of the dataset, it's not accurate when it comes to multiple angles. The lora works best for upper-body shots. On rare occasions it will generate the wrong hair color, just add it to the negative prompt if that ever happens. Some details about training: <ul><li>dataset_count = 48</li><li>optimizer = CAME</li><li>learning_rate = 0.0003</li><li>learning_rate_scheduler = cosine</li><li>epochs = 15</li><li>steps = 720</li><li>lora_rank = 32</li><li>lora_alpha = 1</li><li>batch_size = 1</li></ul>

Files with SHA256: 604711549aded2071ae1ac8b39f4d986dab695db5e42e4d39ca46755eb3431b3