Machines

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

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:

pageMachine Types

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

Gradient makes machines available in the Select a machine section when creating a 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.

When creating a new notebook, we can select from a long list of GPU and CPU-backed machines.

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.

Use the Instance tab on the left side of the notebook to manage compute instances.

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

Select a new instance using the instance selector available on the left side of the notebook.

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.

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.

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.

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

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.

Setting the auto-shutdown timer to 4 hours.

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

The duration remaining on the auto-shutdown timer is visible in the Instance pane.

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

How to view kernel state

By default, all .ipynb files are run on a python3 kernel with a number of preinstalled packages depending on the runtime 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 a programming language specific process that runs independently of other kernels on a virtualized machine. Gradient uses the IPython Jupyter kernel.

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.

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.

Use the Restart Kernel button in the top right to restart a stuck process.

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

It's also possible to use the 3-dot menu to stop or restart a kernel.

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.