<h1 id="description-zll82jgh0">Description</h1>This is a test LoRA model trained on HunyuanVideo using my favorite character.This LoRA depicts the Character Suou Amane from Grisaia series.Trigger word: "suou_amane, red hair, long hair, blue eyes, "Example: "suou_amane, red hair, long hair, blue eyes, an anime girl is looking at viewer. She is wearing school uniform, and smiling to viewer."<h2 id="-v3ti50jkc"></h2><h1 id="usage-notes-8dxik9b2b">Usage Notes</h1>Since this model was trained with a resolution of 960x544, it is recommended to generate videos in similar landscape aspect ratios for optimal results.<h1 id="training-info-f2r09i6z2">Training Info</h1>Used kohya's musubi-tuner for training. <a target="_blank" rel="ugc" href="https://github.com/kohya-ss/musubi-tuner">https://github.com/kohya-ss/musubi-tuner</a>The main training parameters:<ul><li>Learning Rate: 1e-3</li><li>Optimizer Type: adamw8bit</li><li>Network Dim (Rank): 64</li><li>Network Alpha: 1</li><li>Timestep Sampling: sigmoid</li><li>Discrete Flow Shift: 1.0</li><li>Max Train Epochs: 10</li><li>LR Scheduler: cosine_with_min_lr</li><li>LR Scheduler Min LR Ratio: 0.1</li><li>LR Warmup Steps: 0.01</li></ul>Total training steps: 13550<h1 id="dataset-info-elvcbn1fb">Dataset Info</h1>107 images and 230 video clips.The frame extraction settings for videos in musubi-tuner are as follows:<pre><code>frame_extraction = "uniform"
frame_sample = 4
target_frames = [1, 13, 25]</code></pre>

Files with SHA256: 1ae3e8357e11911e2c9d362a4f357d69b3cf206af425809848ba0e4c73b170de