VoltaMLWelcome to VoltaML
Made with ❤️ by Stax124 and Gabe
IMPORTANT
For all Pull Requests, please make sure to target the experimental branch. The main branch is only used for releases (or in some situations, PRs with high priority - marked as Fast-Forward)
INFO
Feel free to join our Discord server
Stable Diffusion WebUI accelerated by AITemplate
This documentation should walk you through the installation process, your first generated image, setting up the project to your liking and accelerating models with AITemplate.
There is also a dedicated section to the Discord bot, API and a section for developers and collaborators.
Main features
- Easy install with Docker
- Clean and simple Web UI
- Supports PyTorch as well as AITemplate for inference
- Support for Windows and Linux
- xFormers support
- Discord bot
- Built-in CivitAI browser
Speed comparison
The below benchmarks have been done for generating a 512x512 image, batch size of one, measured in it/s.
| GPU | PyTorch | SPDA | AITemplate |
|---|---|---|---|
| RTX 4090 | 19 | 39 | 60 |
| RTX 4080 | 15.53 | 20.21 | 40.51 |
| RTX 3070 Laptop | No data | 9.8 | 16.8 |
| RTX 3050 | 4.6 | 5.7 | 10.15 |
| RTX 3060 Ti | No data | 10.50 | 19.46 |
| A100 | 15.1 | 27.5 | No data |
| A10 | 8.8 | 15.6 | 23.5 |
| T4 | 4.3 | 5.5 | No data |
UI Preview
Text to image
Image to image
Image Browser