Want to try out Stable Diffusion but your computer is potato? Don't worry, because this notebook is here to help. "Borrow" Google / Kaggle's servers and run it there.
Disclaimer: This kind of use is currently restricted by Google / Kaggle if you don't have a paid plan. Although no bans are planned, your session may be terminated / deleted at any time. For more information on this, please refer to here and here.
Please note that the Kaggle version may be slightly not up-to-date.
- Open this notebook on Google Colab.
- Log in to your Google account.
- Tweak the options.
- Click
Runtime->Run Allin the menu bar. - (Grant the notebook your Google Drive's access if necessary.)
- Wait until a link pops up in the code output and open it. (it will take a few ten seconds for the link to be reachable.)
- Go to Kaggle.
- Log in to your Kaggle account.
- Click
Create. - In the menu bar, click
File->Import Notebook->GitHub. - Fill in
UjhhgtgTeams/sd-easylized-colabandstable-diffusion-easylized-kaggle-version.ipynb. - Click
Import. - Tweak the options.
- Click
Run Allin the menu bar. - Wait until a link pops up in the code output and open it. (it will take a few ten seconds for the link to be reachable.)
- It takes years for Colab to compile xformers (which is used to speed up image generation) so this procedure is currently disabled.
Solution: You can install it from PyPi for now (available in the config). - Kaggle will not enable GPUs, the Internet, package updates and data persistence for notebooks by default.
Solution: In the bar on the right, look forNotebook Optionsand set the options as below:
Accelerator: GPU T4 * 2
Language: Python
Persistence: Variables and Files
Environment: Always use latest environment
Internet: On
If you couldn't find the options to enable GPUs and the Internet, verify your phone number.
If you would like to help me improve this project or fix bugs, feel free to raise an issue or make a PR. Thanks!