GPU Notebooks
Business Value: Launch fully configured GPU development environments in seconds — no local setup, no driver installation, no environment conflicts.
How It Works
GPU Notebooks provide interactive development environments running directly on Dflare AI's GPU infrastructure. When you launch a notebook:
- The platform provisions a container with your selected GPU allocation
- CUDA runtimes and GPU drivers are pre-installed and validated
- Workspace credentials are auto-injected for seamless access to datasets and MLflow
- Persistent NFS storage ensures your work survives restarts
- Auto-stop policies reclaim idle resources automatically
Technical Highlights
- JupyterLab 4.x and VS Code Server environments
- GPU allocation from fractional GPU to multi-GPU configurations
- Pre-built images with PyTorch, TensorFlow, JAX, and HuggingFace
- Custom container images from any registry (Docker Hub, ECR, GCR)
- MLflow tracking pre-configured — just import and start logging
- Datasets auto-mount at
/data/{dataset-name}
Configuration Options
| Option | Description |
|---|---|
| Environment | JupyterLab or VS Code Server |
| GPU Count | 0 (CPU-only) to 8 GPUs per notebook |
| GPU Type | A100, H100, or other available types |
| Memory | Configurable GPU memory for fractional sharing |
| Auto-Stop | Idle timeout and maximum runtime limits |