Anima Version

This is an experimental LoRA version trained on Anima Preview.

Anima is a new DiT model focused on anime concepts. You can find it on Hugging Face at circlestone-labs/Anima · Hugging Face

Trigger word

@4x0style

Recommended strength is 1.0

Recommendation to Anima Lora Trainer I am working on /ᐠ ̷ ̷𝅒 ̷‸ ̷𝅒 ̷ ᐟ\ﾉ

https://github.com/WalkingMeatAxolotl/AnimaLoraStudio

Experiment using low T and FFL with 1024px and 1536px resolution for capture more texture details.

V1

transformer_path: ~
vae_path: ~
text_encoder_path: ~
t5_tokenizer_path: ~
data_dir: ~
resolution:
- 1024
- 1536
aspect_ratio_limit: 2.0
reg_data_dir: ~
reg_caption: null
reg_weight: 0.5
shuffle_caption: true
keep_tokens: 1
flip_augment: true
tag_dropout: 0.0
prefer_json: true
caption_comfy_encoding: true
cache_latents: true
vae_cache_batch_size: 0
navit_packing: false
cache_encode_tiled: false
lora_type: lora
lora_rank: 64
lora_alpha: 64.0
lora_dora: false
lora_rs: false
lora_dropout: 0.0
lora_rank_dropout: 0.0
lora_module_dropout: 0.0
lora_reg_dims: null
epochs: 40
max_steps: 0
batch_size: 2
grad_checkpoint: true
grad_accum: 2
learning_rate: 8.0e-05
lr_scheduler: none
optimizer_type: adamw
weight_decay: 0.0
kv_trim: false
vae_tiling: auto
noise_enhancement_type: none
timestep_sampling: uniform
timestep_schedule_shift: 0.7
infonoise_enabled: false
loss_type: mse
loss_weighting: none
leap_enabled: false
sra_enabled: false
ffl_enabled: true
ffl_weight: 2.0
ffl_alpha: 1.0
ffl_t_threshold: 1.0
grad_clip_max_norm: 0.0
mixed_precision: bf16