Update README.md
Browse files
README.md
CHANGED
|
@@ -2,16 +2,72 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
|
| 5 |
-
Pre-trained
|
| 6 |
|
| 7 |
Note: The model structure is highly experimental and may be subject to change in the future.
|
| 8 |
|
| 9 |
Inference with ComfyUI: https://github.com/kohya-ss/ControlNet-LLLite-ComfyUI
|
| 10 |
|
|
|
|
|
|
|
| 11 |
Training: https://github.com/kohya-ss/sd-scripts/blob/sdxl/docs/train_lllite_README.md
|
| 12 |
|
| 13 |
The recommended preprocessing for the blur model is Gaussian blur.
|
| 14 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
|
| 16 |
|
| 17 |
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
|
| 5 |
+
Pre-trained models and output samples of ControlNet-LLLite.
|
| 6 |
|
| 7 |
Note: The model structure is highly experimental and may be subject to change in the future.
|
| 8 |
|
| 9 |
Inference with ComfyUI: https://github.com/kohya-ss/ControlNet-LLLite-ComfyUI
|
| 10 |
|
| 11 |
+
For 1111's Web UI, [sd-webui-controlnet](https://github.com/Mikubill/sd-webui-controlnet) extension supports ControlNet-LLLite.
|
| 12 |
+
|
| 13 |
Training: https://github.com/kohya-ss/sd-scripts/blob/sdxl/docs/train_lllite_README.md
|
| 14 |
|
| 15 |
The recommended preprocessing for the blur model is Gaussian blur.
|
| 16 |
|
| 17 |
+
# Naming Rules
|
| 18 |
+
|
| 19 |
+
`controllllite_v01032064e_sdxl_blur_500-1000.safetensors`
|
| 20 |
+
|
| 21 |
+
- `v01` : Version Flag.
|
| 22 |
+
- `032` : Dimensions of conditioning.
|
| 23 |
+
- `064` : Dimensions of control module.
|
| 24 |
+
- `sdxl` : Base Model.
|
| 25 |
+
- `blur` : The control method. `anime` means the LLLite model is trained on/with anime sdxl model and images.
|
| 26 |
+
- `500-1000` : (Optional) Timesteps for training. If this is `500-1000`, please control only the first half step.
|
| 27 |
+
|
| 28 |
+
# Models
|
| 29 |
+
|
| 30 |
+
## Trained on sdxl base
|
| 31 |
+
|
| 32 |
+
- controllllite_v01032064e_sdxl_blur-500-1000.safetensors
|
| 33 |
+
- trained with 3,919 generated images and Gaussian blur preprocessing.
|
| 34 |
+
- controllllite_v01032064e_sdxl_canny.safetensors
|
| 35 |
+
- trained with 3,919 generated images and canny preprocessing.
|
| 36 |
+
- controllllite_v01032064e_sdxl_depth_500-1000.safetensors
|
| 37 |
+
- trained with 3,919 generated images and MiDaS v3 - Large preprocessing.
|
| 38 |
+
|
| 39 |
+
## Trained on anime model
|
| 40 |
+
|
| 41 |
+
The model ControlNet trained on is our custom model.
|
| 42 |
+
|
| 43 |
+
- controllllite_v01016032e_sdxl_blur_anime_beta.safetensors
|
| 44 |
+
- beta version.
|
| 45 |
+
- controllllite_v01032064e_sdxl_blur-anime_500-1000.safetensors
|
| 46 |
+
- trained with 2,836 generated images and Gaussian blur preprocessing.
|
| 47 |
+
- controllllite_v01032064e_sdxl_canny_anime.safetensors
|
| 48 |
+
- trained with 921 generated images and canny preprocessing.
|
| 49 |
+
- controllllite_v01008016e_sdxl_depth_anime.safetensors
|
| 50 |
+
- trained with 1,433 generated images and MiDaS v3 - Large preprocessing.
|
| 51 |
+
- controllllite_v01032064e_sdxl_fake_scribble_anime.safetensors
|
| 52 |
+
- trained with 921 generated images and PiDiNet preprocessing.
|
| 53 |
+
- controllllite_v01032064e_sdxl_pose_anime.safetensors
|
| 54 |
+
- trained with 921 generated images and MMPose preprocessing.
|
| 55 |
+
- controllllite_v01032064e_sdxl_pose_anime_v2_500-1000.safetensors
|
| 56 |
+
- trained with 1,415 generated images and MMPose preprocessing.
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
# Samples
|
| 60 |
+
|
| 61 |
+
## sdxl base
|
| 62 |
+
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
|
| 66 |
+
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
## anime model
|
| 70 |
+
|
| 71 |
|
| 72 |
|
| 73 |
|