This Docker image focuses on secure training and inference with Stable Diffusion by securely storing and fetching datasets and checkpoints from the Storj Decentralized Cloud Storage.
- RunPod: https://runpod.io/gsc?template=3jgnr4mriw&ref=se98m3j8
- VALDI: Start and connect to a Direct VM
- Stable Diffusion Web UI 1.5.1
- Stable Studio
- Automated fetching of AI models from your Storj account on startup
- Jupyter Notebook for DreamBooth training and storing the trained checkpoints to your Storj account.
| Var name | Description |
|---|---|
| STORJ_ACCESS | The access grant to use for transferring files from and to your Storj DCS account |
| STORJ_BUCKET | The bucket to use for transferring files from and to your Storj DCS account |
| STORJ_MODEL_PATH | The object key to a specific AI model or a prefix/folder with multiple AI models to automatically download at startup |
If the Docker container is configured with all three environment variables, it will automatically download the AI models at startup.
The usage instructions are maintained in the wiki.
Instructions for RunPod
- Deploy and launch a pod
- Configure the pod to download your AI models from Storj automatically
- Launch the Stable Diffusion Web UI
- Launch the Stable Studio
- Train your model and upload it to Storj