Gradient Docs
Gradient HomeHelp DeskCommunitySign up free
Gradient Next
Gradient Next
  • About Gradient
  • Get Started
    • Quick Start
      • Install the Gradient CLI
    • Core Concepts
    • Organizing Projects
      • Secrets
      • Storing an API key as a Secret
    • Tutorials
      • Gradient Notebooks Tutorial
      • Gradient Workflows Tutorial
      • Gradient Deployments Tutorial
    • FAQ
    • Common Errors
  • Gradient Platform
    • Gradient Notebooks
      • Runtimes
      • Files and storage
      • Machines
      • Terminal
      • Shortcuts
      • Sharing
      • TensorBoard
      • Run on Gradient
    • Gradient Workflows
      • Basic operations
      • Understanding Inputs & Outputs
      • Workflow Spec
      • Gradient Actions
      • Environment Variables
      • Using YAML for Data Science
    • Gradient Deployments
      • Basic operations
      • Deployment Spec
  • Artifacts
    • Container Management
      • Custom Containers
    • Data
      • Versioned Data
        • Public Datasets Repository
        • Storage Providers
      • Persistent Storage
    • Models
      • Managing Models
        • Model Types & Metadata
        • Public Models
    • Code
    • Metrics
      • Push Metrics
      • View & Query Metrics
  • Gradient Cluster
    • Overview
      • Setup
        • Managed Private Clusters
        • Self-Hosted Clusters
          • Pre-installation steps
          • Gradient Installer CLI
          • Terraform
            • Pre-installation steps
            • Install on AWS
            • Install on bare metal / VMs
            • Install on NVIDIA DGX
          • Let's Encrypt DNS Providers
          • Updating your cluster
      • Usage
  • More
    • SDK
      • Projects Client
      • Models Client
      • Deployments Client
      • Workflows Client
      • SDK Examples
      • Full SDK Reference
    • Machine Types
      • Machine Tiers
      • Free Machines (Free Tier)
    • Your Account
      • Teams
        • Creating a Team
        • Upgrading to a Team Plan
      • Hotkeys
      • Billing & Subscriptions
        • Storage Billing
      • Public Profiles
    • Release notes
    • Roadmap
Powered by GitBook
On this page
  • Introduction to notebook machines
  • How to select a machine when creating a notebook
  • How to swap machines in an existing notebook
  • How to use a machine that is unavailable
  • How to use the auto-shutdown timer
  • How to view kernel state
  • How to restart a kernel
  1. Gradient Platform
  2. Gradient Notebooks

Machines

This guide explains how instances and kernels are used in Gradient Notebooks

PreviousFiles and storageNextTerminal

Last updated 3 years ago

Introduction to notebook machines

Gradient makes a wide variety of machines available for use with notebooks. A full list of instance types provided by Gradient is available here:

The first opportunity to select a machine is in the Select a machine section when creating a new notebook.

Gradient Notebooks makes it easy to swap out a machine for a different machine at any time -- so keep that in mind when deciding what machine to select. It's common to start with a less powerful machine and upgrade as needed over time.

How to select a machine when creating a notebook

When creating a new notebook, the Select a machine section of the notebook create workflow provides a list of GPU-backed and CPU-backed machines available for use.

If a machine is available for use it will be listed under the Available heading. If a machine is out of capacity it means that all instances are currently being used by other Paperspace users.

How to swap machines in an existing notebook

To swap instances in an existing notebook, use the Instance selector available on the left side of the notebook.

If the notebook is currently in the Running state, it will first need to be stopped.

After the instance is stopped, a new instance can be selected from the list of available machines.

The top bar will indicate when a notebook is again in the Running state.

How to use a machine that is unavailable

If an instance type is at capacity it will be listed as Unavailable. Popular instance types, such as instances providing free gpus, sometimes reach capacity during busy periods.

If the machine type that we require is listed as unavailable, we'll need to wait until capacity frees up. We should first try waiting a few minutes and refreshing the page. If capacity is still not available, we might try a different instance with better availability or come back during a less busy time of day.

Free GPU instances in particular are often in high demand. Try using popular free resources outside of peak hours (MF 9:00AM - 5:00PM ET) or try using more powerful paid machines.

How to use the auto-shutdown timer

The auto-shutdown timer is a useful feature that automatically shuts down a notebook after a pre-determined amount of time. The timer starts from the moment that the notebook is started.

When a notebook is running, the amount of time remaining until auto-shutdown is displayed in the Instance tab.

It is currently not possible to alter the auto-shutdown interval after it has been set.

How to view kernel state

Each time a .ipynb file is created, uploaded or run,a new kernel is created to manage that file. The kernel state is visible in the file manager as well as in the top bar.

The kernel state is made accessible in the top bar as text. In the image above, for example, the top bar is indicating that the kernel state is Running.

How to restart a kernel

One of the benefits of kernels is the ability to restart a single kernel without disrupting the rest of the notebook. This is useful when a notebook is stuck or unresponsive.

An individual kernel may be restarted using the Restart Kernel button in the top right of the notebooks IDE.

It's also possible to stop or restart the kernel using the 3-dot menu next to an individual file in the file manager.

If restarting the kernel does not fix an issue with a stuck process, it may be necessary to stop and restart the entire notebook. In this case we would select Stop Instance and wait for the instance to shutdown before selecting a new instance to start.

Paperspace is constantly adding capacity to its datacenters. The is a good resource to hear about capacity upgrades. If you are repeatedly running into capacity issues with a particular machine type, please .

By default, all .ipynb files are run on a python3 kernel with a number of preinstalled packages depending on the you choose. This is meant to provide an environment that works right out of the box for the majority of projects.

Gradient allows the user to start an arbitrary number of kernels within the notebooks IDE. A kernel is that runs independently of other kernels on a virtualized machine. Gradient uses the Jupyter kernel.

Changelog
let us know
runtime
a programming language specific process
IPython
Machine Types
Gradient makes machines available in the Select a machine section when creating a notebook.
When creating a new notebook, we can select from a long list of GPU and CPU-backed machines.
Use the Instance tab on the left side of the notebook to manage compute instances.
Select a new instance using the instance selector available on the left side of the notebook.
When machines are at capacity, they are marked as Currently Unavailable. It's recommended to wait a few minutes and check back to see if capacity has opened up or else to try a different machine type that is available.
Setting the auto-shutdown timer to 4 hours.
The duration remaining on the auto-shutdown timer is visible in the Instance pane.
Kernel state is visible in the file manager as well as in the top bar.
Use the Restart Kernel button in the top right to restart a stuck process.
It's also possible to use the 3-dot menu to stop or restart a kernel.