Type | |
Stats | 188 12,843 |
Reviews | (101) |
Published | Jul 25, 2024 |
Base Model | |
Training | Steps: 900 Epochs: 15 |
Usage Tips | Clip Skip: 2 |
Trigger Words | orange peel |
Hash | AutoV2 11FFE1C43A |
This is a preliminary experimental Lora model based on orange peel textures. While it performs well on certain objects, it may not always function as intended on others, such as biological entities or vehicles. Compatibility can be a bit of a hit or miss. Perhaps I'll refine it further in the future to improve its compatibility.
🙇♀️ Apology Notice🙇
After numerous attempts, it's clear that my Lora has several flaws. It struggles to accurately generate specific objects, characters, and creatures. Setting a high weight often results in the desired elements not being generated, while a lower weight leads to less pronounced orange peel textures.
Additionally, watch straps frequently appear, and there's a tendency for strange objects to be included. When generating objects, human figures often intrude. Furthermore, compatibility with certain Pony checkpoints is subpar.
As this was my first attempt at training a Pony model, I didn't conduct thorough testing, leading to these significant issues. I apologize for inconveniencing everyone during the Daily Challenges with this unsatisfactory model.