<h2>Pi (WILL: A Wonderful World) Pi (WILL：美好世界)</h2><a target="_blank" rel="ugc" href="https://civarchive.com/models/38103/experimentalalicia-will-a-wonderful-world-will">Alicia (WILL: A Wonderful World) 阿莉西亚 (WILL：美好世界)</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/37821">Li Wen (WILL: A Wonderful World) 李雯 (WILL：美好世界)</a>Triggers with <code>pi_will</code>, weight has to be 0.6-0.8.This LoRA is an experiment to answer whether I can make a decent LoRA with only copies of the same image. I left two different epochs for you guys to download and test.For Pi, the dataset was like this:<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/39d1f6c3-33ee-4dd9-eee8-29ce8f55fa00/width=525/39d1f6c3-33ee-4dd9-eee8-29ce8f55fa00.jpeg" />Used Cetus-mix and Anypastel for example images.<h2>Usage</h2>General tags: <code>pi_will, pi, 1boy, bishounen, short hair, blonde hair, blue eyes, mole under eye</code>For his casual clothes, just use <code>default_outfit, blue coat, white sweater, black pants, stud earring</code>.<h2>Known Issues</h2><ul><li>Don't bother using it at a weight higher than 0.8, it'll overfit.</li><li>Sometimes his sweater will appear as black.</li></ul><h2>Training Parameters</h2><ul><li>"ss_text_encoder_lr":"1e-05",</li><li>"ss_total_batch_size":"2",</li><li>"ss_num_epochs":"10",</li><li>"ss_unet_lr":"0.0001",</li><li>"ss_learning_rate":"0.0001",</li><li>"ss_lr_scheduler":"cosine_with_restarts",</li><li>"ss_lr_warmup_steps":"0",</li><li>"ss_max_token_length":"225",</li><li>"ss_mixed_precision":"fp16",</li><li>"ss_network_alpha":"64.0",</li><li>"ss_network_dim":"128",</li><li>{\"n_repeats\": 20, \"img_count\": 10}</li><li>"ss_resolution":"(768, 768)"</li><li>"ss_sd_model_name":"Animefull-final-pruned.ckpt",</li><li>"ss_seed":"42",</li></ul>

Files with SHA256: bfa4921863139bb81f2a46c77492b87719b8488d02e6c713ca992fbfb813f512