|
--- |
|
tags: |
|
- stable-diffusion |
|
pipeline_tag: text-to-image |
|
license: apache-2.0 |
|
--- |
|
|
|
# SDXL-PD6-Character-Dora-1024 |
|
|
|
A collection of various character loras (some of you may be familiar with my posts on [CivitAI](https://civitai.com/user/DocShotgun)) trained at 1024 resolution on [Pony Diffusion V6 XL](https://civitai.com/models/257749) and intended to be used with said model as well as derivative models such as [AutismMix SDXL](https://civitai.com/models/288584). |
|
|
|
These are trained using [DoRA](https://arxiv.org/abs/2402.09353) targetting the linear and conv layers of the unet while freezing the text encoder. All datasets were re-labeled and some were expanded with additional images to improve quality. |
|
|
|
Training was done using manually-gathered datasets and [Derrian Distro's training scripts](https://github.com/derrian-distro/LoRA_Easy_Training_Scripts) on 1x 4090. Detailed training data, including trained tags, is provided in the safetensors metadata included with each lora. |
|
|
|
These are provided as-is, and YMMV. The user is responsible for any outputs produced using these loras, and outputs may vary depending on the base model used. |
