<h3 id="heading-972">概况</h3>基于<a rel="ugc" href="https://civarchive.com/models/260267/animagine-xl-v3">animagineXL v3</a>和<a rel="ugc" href="https://civarchive.com/models/332076/kohaku-xl-delta">kohaku delta</a>训练的blue archive的plana人物lokr模型，也能用于出arona。<ul><li>触发词plana\(blue archive\),采用详细标注的方式打标训练，因此想要完整还原人物特征时需要尽可能的添加相关特征的提示词；</li><li>由于训练集本身混有一定arona的图片，因此也能用arona (blue archive)来出arona，但是效果上没有那么优秀；</li><li>animagineXL v3这个模型（kohaku delta也能，但是效果稍弱）本身就能使用提示词直出plana和arona，这个lokr旨在通过训练进一步还原这两个人物的人物特征；</li><li>初步测试未观察到明显影响画风，可考虑与画风/艺术家提示词或者是画风lora一起使用；</li><li>未测试这两个版本的lokr在非对应底模的迁移性。</li></ul><h3 id="heading-932">训练参数</h3>遵循<a target="_blank" rel="ugc" href="https://freedevproject.org/faipl-1.0-sd/">Fair AI Public License 1.0-SD</a>协议,公开训练参数,在<a rel="ugc" href="https://github.com/kohya-ss/sd-scripts">kohya-ss</a>的训练脚本上进行的训练。<ul><li>for an3xl v1.0版本，训练集共647张图片。</li></ul><pre><code>pretrained_model_name = "animagineXLV3_v30.safetensors"
learning_rate = 6e-06
resolution = "1024,1024"
lr_scheduler = "cosine"
train_batch_size = "4"
gradient_accumulation_steps = 4
max_train_steps = "1200"
mixed_precision = "fp16"
seed = "1919810"
optimizer_type = "lion8bit"
network_module = lycoris.kohya
network_args "preset=attn-mlp" "factor=8" "algo=lokr"
network_dim = 100000
multires_noise_iterations = "8" 
multires_noise_discount = "0.6" 
network_train_unet_only = true
min_snr_gamma = 10
enable_bucket = true
bucket_reso_steps = 32
min_bucket_reso = 512
max_bucket_reso = 2048
class_tokens = 'plana\(blue archive\)'
shuffle_caption = true
gradient_checkpointing = true
xformers = true
cache_latents_to_disk = true
persistent_data_loader_workers = true</code></pre><ul><li>for delta v1.0版本，训练集共722张图片。</li></ul><pre><code>
pretrained_model_name = "kohakuXLDelta_rev1.safetensors"
learning_rate = 1e-05
resolution = "1024,1024"
lr_scheduler = "cosine"
train_batch_size = "4"
gradient_accumulation_steps = 4
max_train_steps = "1200"
mixed_precision = "fp16"
seed = "1919810"
optimizer_type = "lion8bit"
network_module = lycoris.kohya
network_args "preset=attn-mlp" "factor=8" "algo=lokr"
network_dim = 100000
multires_noise_iterations = "8" 
multires_noise_discount = "0.6" 
network_train_unet_only = true
min_snr_gamma = 10
enable_bucket = true
bucket_reso_steps = 32
min_bucket_reso = 512
max_bucket_reso = 2048
class_tokens = 'plana\(blue archive\)'
shuffle_caption = true
gradient_checkpointing = true
xformers = true
cache_latents_to_disk = true
persistent_data_loader_workers = true</code></pre><h3 id="heading-973">执照</h3><a target="_blank" rel="ugc" href="https://freedevproject.org/faipl-1.0-sd/">Fair AI Public License 1.0-SD</a>

plana (blue archive)[lokr]

基于kohaku delta训练的plana人物lokr模型

Files with SHA256: 8bc9bb5a1b404ab9c032323d4bb27e452656d454cd3580860cde039739aa71e6