<h2 id="monody-illustrious-fuesie89p">Monody-Illustrious</h2>This model focuses on making the character look cute.The datasets used were carefully selected manually one by one from all over the internet to suit the objectives of this model.No artist tags are used during training because this model aims only to change the character's appearance.<h2 id="update:-version-a1-udtl4p2a0">Update: version A1</h2>This version is an improvement from the previous version by merging 2 LoRas which were trained to focus on improving the anatomy and color of the previous model.The downside of combining LoRa is that the artist style and some of the characters that have been trained won't really work in this modelYou can look in the about this version section to see the details of the Merge Lora.For Testing :<a target="_blank" rel="ugc" href="https://tensor.art/models/806986757045386027/Monody-illustrious-A1">TensorArt</a><a rel="ugc" href="https://www.shakker.ai/modelinfo/906a473857f1428e8e39848a76341132/Monody-Illustrious?mine=1&amp;from=personal_page">Shakker</a>

Monody-Illustrious

Training Detail :GPU : Nvidia L40Slearning_rate = 5e-6train_text_encoder = truetrain_batch_size = 4gradient_accumulation_steps = 2optimizer_type = "AdamW8bit"lr_scheduler = "CosineAnnealingWarmupRestarts""weight_decay=0.1", "betas=0.9,0.99"

Files with SHA256: 1c1885b2584ac4e62b3a5505928defba225aab84f27f07457b4bb3849d140fae