Kuchiki Rukia - BleachOnly trained on Season 1 with <a target="_blank" rel="ugc" href="https://github.com/negaga53/neme-anima">https://github.com/negaga53/neme-anima</a> on Anima Base v1.Only trained on images without post processing (bright/dark scenes, flashbacks, etc.)<h3 id="triggers">=== Triggers ===</h3>Character: rkia, official style, kuchiki rukia,Hair/Eyes: black hair, hair between eyes, purple eyesOutfit 1: black kimono, black hakama, sandals, // katanaOutfit 2: blazer, school uniform, skirt, kneehighs, bow, ribbonOutift 3: white shirt, school uniform, skirt, kneehighs, bow, ribbon Extra: pajamasPreview images have been upcsaled with x2 Hires Fix and (partially) with Adetailer (Face and Hand).Use strength 0.8 - 1.0This was one of the things I REALLY missed from Illustrious and I finally beat my laziness and trained it myself. It's not perfect by no means but it gets the job done pretty well.<h2 id="if-you-think-you-can-create-a-better-lora-please-do-it!">If you think you can create a better lora, PLEASE do it! 😭</h2><h2 id="i-attached-the-training-data.">I attached the training data. </h2><h3 id="technical-details">=== Technical details ===</h3>Training parameters from run.toml and dataset.toml:epochs = 60micro_batch_size_per_gpu = 1pipeline_stages = 1gradient_accumulation_steps = 2gradient_clipping = 1.0warmup_steps = 100eval_every_n_epochs = 5eval_before_first_step = trueeval_micro_batch_size_per_gpu = 1eval_gradient_accumulation_steps = 1save_every_n_epochs = 10activation_checkpointing = truepartition_method = "parameters"save_dtype = "bfloat16"caching_batch_size = 1map_num_proc = 8steps_per_print = 1compile = true[model]type = "anima"transformer_path = "/home/main/.cache/neme-anima/models/anima-base-v1.0.safetensors"vae_path = "/home/main/.cache/neme-anima/models/qwen_image_vae.safetensors"llm_path = "/home/main/.cache/neme-anima/models/qwen_3_06b_base.safetensors"dtype = "bfloat16"llm_adapter_lr = 0.0sigmoid_scale = 1.3[adapter]type = "lora"rank = 32dtype = "bfloat16"[optimizer]type = "adamw_optimi"lr = 5e-05betas = [0.9, 0.99]weight_decay = 0.01eps = 1e-08resolutions = [512, 1024, 1536]enable_ar_bucket = truemin_ar = 0.5max_ar = 2.0num_ar_buckets = 9num_repeats = 1.0

Files with SHA256: 60aa9ea2db9d24c7afbed891598dc285660d1ca9e40186f7385c9e5e79eed245