Type | |
Stats | 181 0 |
Reviews | (10) |
Published | Apr 5, 2024 |
Base Model | |
Training | Steps: 2,047 Epochs: 89 |
Trigger Words | lilyCole01 |
Hash | AutoV2 EB782BEA3D |
Lily Luahana Cole is a British model, actress, director, author and humanitarian. Born in 1987, she's best-known for her work with the fashion magazine Vogue, as well as her lead role in the film The Imaginarium of Doctor Parnassus.
The embedding training dataset consists of 150 publicly-available photos from the early-mid 2010s, cropped and masked for learning attention using the amazing OneTrainer. With repeats, there were 230 images per epoch, processed in batches of 10 (23 gradient accumulation steps per epoch) for 89 epochs. The ADAMW optimizer was utilized with a learning rate of 0.00075 and 500 LR warmup steps. The dataset contains a fair quantity of pinup photos, so generation results may require appropriate negative prompting to prevent NSFW outputs, depending on model leanings. Please be respectful in your prompts.
I'd love to see some creations made with this Ti. Happy prompting!
SHA256 hash: EB782BEA3DA0BC3AD2DCB761CB022A82FD10EC251BA323FEBA4CBD951F17680E