Paperspace

Paperspace (now part of DigitalOcean) provides GPU-powered cloud machines and managed Jupyter notebooks for ML experimentation, training, and deployment. It offers two products: Gradient (a managed ML platform with notebooks, workflows, and deployments) and Core (raw GPU virtual machines you SSH into).

• GPU notebooks — launch Jupyter notebooks with GPU acceleration in seconds, with pre-built ML runtimes for PyTorch, TensorFlow, and JAX
• Gradient workflows — define reproducible ML pipelines as YAML that run on GPU-backed containers with versioned datasets and experiments
• GPU virtual machines — spin up full Linux VMs with A100, A6000, or RTX GPUs for custom training setups and long-running experiments
• Model deployment — deploy trained models as API endpoints with Gradient Deployments, including auto-scaling and versioning
• Persistent storage — attach shared storage volumes to notebooks and jobs for datasets and model checkpoints
• Free GPU tier — access free M4000 and P5000 GPU notebooks for learning and small experiments

Individual ML practitioners, students, researchers, and small teams choose Paperspace for its approachable pricing and user-friendly interface. It occupies a sweet spot between local development and enterprise cloud platforms — more capable than Google Colab, simpler than AWS SageMaker.

Paperspace is particularly popular with the fast.ai community and ML learners who want affordable GPU access for courses, tutorials, and personal projects without committing to enterprise-grade (and enterprise-priced) cloud platforms.

1. Create an account — sign up at paperspace.com. Choose between Gradient (managed ML) or Core (raw VMs).
2. Launch a Gradient notebook — click "Create Notebook," select a GPU machine type (Free GPU, P5000, A100), choose a pre-built runtime (PyTorch, TensorFlow, or custom), and start coding immediately in the browser.
3. Or launch a Core VM — select a GPU machine type from the Core dashboard, choose an OS template (Ubuntu with ML drivers pre-installed), and SSH in:

   ssh paperspace@your-vm-ip

4. Use Gradient CLI — for programmatic access, install the CLI:

   pip install gradient
   gradient apiKey YOUR_API_KEY

   Create notebooks and run workflows from the command line:

   gradient notebooks create --machineType P5000 --container tensorflow/tensorflow:latest-gpu

5. Deploy a model — use Gradient Deployments to serve models as APIs with auto-scaling and versioning from the dashboard or CLI.

Pricing: Free tier includes GPU notebooks (M4000, limited hours). P5000: ~$0.51/hr. A100 80GB: ~$3.09/hr. Core VMs billed hourly or monthly. Persistent storage starts at $0.29/GB/month. Full pricing details.