<h2 id="turbo-lora-distillation-of-kirazuri-(anima)-v4.0">Turbo LoRA distillation of <a target="_blank" rel="ugc" href="https://civarchive.com/models/2495369/kirazuri-anima?modelVersionId=3112659">Kirazuri (Anima) v4.0</a> </h2>Important: This LoRA is not expected to work with other base models. Generation Settings:4 steps, CFG 1sampler/scheduler:<code>res_2s</code> + <code>bong_tangent</code><code>res_multistep</code> + <code>simple</code><code>dpmpp_2m</code> + <code>sgm_uniform</code>Some artifacting is noticeable with samplers that work better at higher step counts i.e. eulerTried to address this a somewhat in version 2.0, and these scheduler combinations also appear work much better at the low 4 step count.Otherwise, please refer to recommendations on the <a target="_blank" rel="ugc" href="https://civarchive.com/models/2495369/kirazuri-anima?modelVersionId=3112659">Kirazuri (Anima) </a>base model card.Training details:Created with DP-DMD diversity-preserved few-step distillation.Thanks to sorryhyun for the trainer: <a target="_blank" rel="ugc" href="https://github.com/sorryhyun/anima_lora">https://github.com/sorryhyun/anima_lora</a>Dataset: <a target="_blank" rel="ugc" href="https://civarchive.com/models/929497/aesthetic-quality-modifiers-masterpiece">https://civarchive.com/models/929497/aesthetic-quality-modifiers-masterpiece</a>

Files with SHA256: e5c0d6d29165e0afc1e42f35f7a8ef706ff8afe41c054b6fa4cf924621fa1fc8