<h3 id="!!!this-is-still-a-work-in-progress!!!">!!!THIS IS STILL A WORK IN PROGRESS!!!</h3><h1 id="heeeeyyyy!">Heeeeyyyy!</h1>For those of you that do not know. She is an artic wolf who starts out as a spoiled brat that cannot act but can sing good.<h2 id="tips-and-tricks">Tips and Tricks</h2><ul><li>If you get a human or partial, try adding "wolf girl" or "animated wolf girl".</li><li>Using "paws" instead of "hands" can make her hands more accurate. (even though I trained using "handpaws"</li></ul><h2 id="sampler-and-scheduler-combos">Sampler and Scheduler Combos</h2><ul><li>dpmpp_2m_sde &amp; beta</li><li>dpmpp_2m_sde &amp; simple</li><li>dpmpp_sde &amp; beta</li><li>dpmpp_sde &amp; simple</li><li>res_multistep &amp; simple</li></ul><h3></h3>Anyways, I have spent a good while trying to fine tune the dataset and the training parameters.I am open to any suggestions on improving the model.If you create an image of Porsha that looks movie accurate, I might add it to the dataset and toss some yellow buzz your way.This cannot be used for any commercial purposes since I do not own the character. This is purely for fun. I have only posted this on Civitai.

Porsha Crystal - Sing 2 - ZIT

Trained using Ostris AI Toolkit: linear: 32 linear_alpha: 32 conv: 16 conv_alpha: 16 lokr_full_rank: true lokr_factor: -1 network_kwargs: ignore_if_contains: [] save: dtype: bf16 save_every: 250 max_step_saves_to_keep: 4 save_format: diffusers datasets: mask_min_value: 0.1 caption_dropout_rate: 0.05 cache_latents_to_disk: false is_reg: false network_weight: 1 resolution: - 512 - 768 - 1024 controls: shrink_video_to_frames: true num_frames: 1 do_i2v: true train: batch_size: 1 bypass_guidance_embedding: false steps: 2800 gradient_accumulation: 1 train_unet: true train_text_encoder: false gradient_checkpointing: true noise_scheduler: flowmatch optimizer: adamw8bit timestep_type: weighted content_or_style: balanced optimizer_params: weight_decay: 0.01 unload_text_encoder: false cache_text_embeddings: false lr: 0.0001 ema_config: use_ema: true ema_decay: 0.99 skip_first_sample: true force_first_sample: false disable_sampling: false dtype: bf16 diff_output_preservation: false diff_output_preservation_multiplier: 1 diff_output_preservation_class: person switch_boundary_every: 1 loss_type: mse model: name_or_path: Tongyi-MAI/Z-Image-Turbo quantize: true qtype: qfloat8 quantize_te: true qtype_te: qfloat8 arch: zimage:turbo low_vram: true layer_offloading: false layer_offloading_text_encoder_percent: 1 layer_offloading_transformer_percent: 1

Files with SHA256: bf5fd6aebf6a35ea085923c198c9884e2729d54d219b75b0b0e4eb78700f4a26