https://github.com/replicate/cog-sdxl
Stable Diffusion XL training and inference as a cog model
https://github.com/replicate/cog-sdxl
Last synced: 9 months ago
JSON representation
Stable Diffusion XL training and inference as a cog model
- Host: GitHub
- URL: https://github.com/replicate/cog-sdxl
- Owner: replicate
- License: apache-2.0
- Created: 2023-08-01T18:50:46.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2024-11-08T22:25:25.000Z (about 1 year ago)
- Last Synced: 2025-05-16T01:34:24.517Z (9 months ago)
- Language: Python
- Homepage: https://replicate.com/stability-ai/sdxl
- Size: 27.9 MB
- Stars: 224
- Watchers: 24
- Forks: 107
- Open Issues: 34
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Cog-SDXL
[](https://replicate.com/stability-ai/sdxl)
This is an implementation of Stability AI's [SDXL](https://github.com/Stability-AI/generative-models) as a [Cog](https://github.com/replicate/cog) model.
## Development
Follow the [model pushing guide](https://replicate.com/docs/guides/push-a-model) to push your own fork of SDXL to [Replicate](https://replicate.com).
## Basic Usage
for prediction,
```bash
cog predict -i prompt="a photo of TOK"
```
```bash
cog train -i input_images=@example_datasets/__data.zip -i use_face_detection_instead=True
```
```bash
cog run -p 5000 python -m cog.server.http
```
## Update notes
**2023-08-17**
* ROI problem is fixed.
* Now BLIP caption_prefix does not interfere with BLIP captioner.
**2023-08-12**
* Input types are inferred from input name extensions, or from the `input_images_filetype` argument
* Preprocssing are now done with fp16, and if no mask is found, the model will use the whole image
**2023-08-11**
* Default to 768x768 resolution training
* Rank as argument now, default to 32
* Now uses Swin2SR `caidas/swin2SR-realworld-sr-x4-64-bsrgan-psnr` as default, and will upscale + downscale to 768x768