Verified: a year ago
SafeTensor
Type | |
Stats | 385 4,198 |
Reviews | (72) |
Published | Jan 20, 2024 |
Base Model | |
Training | Steps: 2,700 Epochs: 10 |
Usage Tips | Clip Skip: 1 |
Trigger Words | swampert |
Hash | AutoV2 E9AF18E38D |
Swampert LoRA trained on NovelAI with Kohya training scripts using the DAdaptAdam optimizer at 512x512 resolution. Works great, but has issues. Tail placement and color are inconsistent. Placement of the orange 'pods' on the arms and legs are inconsistent. Head fins can try to become more 'ear' like depending on the base model used. Character detail is weakened when prompting for clothing, likely due to the lack of clothed images in the training data. Increasing LoRA strength should help with clothed images, but can cause other issues. Should have wide compatibility with model mixes based on NovelAI, which is most of them. Trained on a mixed male/female dataset.