CivArchive
    [Style] Artist: 龍犬 | Ryuudog - v1.0
    NSFW
    Preview 145634494
    Preview 145634493
    Preview 145634499
    Preview 145634501

    Introduction

    This is a Ryuudog style LoRA trained on Anima Base 1.0.

    Usage

    Recommended LoRA strength: 1.0
    Trigger word: @ryuudog

    Training Details

    • Base model: Anima Base 1.0

    • Trainer: sorryhyun/anima_lora

    • Dataset: 111 artworks by the artist, including 25 selected artworks and 86 other artworks

    • Sampling weights: 3:1 for the selected artworks versus the remaining artworks

    • LoRA rank / alpha: 32 / 128

    • AdaLN rank: 16

    • Learning rate: 2e-5

    • Optimizer: AdamW

    • Scheduler: Cosine, with 5% warmup

    • Batch size: 1

    • Training duration: 2,220 steps

    • Precision: BF16

    • Resolution: Native 1024/896 aspect-ratio buckets, approximately 1 megapixel

    • REPA: Relational REPA, weight 0.05, layer 8

    • GPU: 1 × NVIDIA RTX 3090

    简介

    这是基于 Anima Base 1.0 训练的 Ryuudog 画风 LoRA。

    使用

    推荐 LoRA 强度: 1.0
    激活词: @ryuudog

    训练参数

    • 基础模型: Anima Base 1.0

    • 训练工具: sorryhyun/anima_lora

    • 数据集: 111 张画师作品,其中精选作品 25 张,其余作品 86 张

    • 采样权重: 精选作品与其余作品为 3:1

    • LoRA rank / alpha: 32 / 128

    • AdaLN rank: 16

    • 学习率: 2e-5

    • 优化器: AdamW

    • 学习率调度: Cosine,5% warmup

    • Batch size: 1

    • 训练量: 2,220 steps

    • 训练精度: BF16

    • 训练分辨率: 原生 1024/896 宽高比分桶,约 1 M 像素

    • REPA: Relational REPA,权重 0.05,第 8 层

    • 显卡: 1 × NVIDIA RTX 3090

    Description

    LORA
    Anima

    Details

    Downloads
    80
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/14/2026
    Updated
    10/14/2026
    Deleted
    -
    Trigger Words:
    @ryuudog

    Files

    ryuudog-best3-step2220.safetensors

    Mirrors