Sometimes you just need a big machine in the cloud.
You can use Coiled from anywhere you run Python. You do not need to use Coiled clusters from Coiled notebooks.
This feature is experimental and may be removed or changed without notice.
You can easily launch Jupyter servers running in the cloud with many of the same advantages you get from Coiled:
The same package versions you run locally
The same cloud permissions you have locally
The same files you have locally with the
--synckeyword (unique to notebooks)
In any region on AWS/GCP
With robust logging and metrics collection
With no additional cloud infrastructure (like Kubernetes)
You can ask for a notebook as follows:
coiled notebook start
This will take 1–3 minutes and open a browser tab for you connecting you to an authenticated Jupyter server. You can shut down that server as follows:
coiled notebook stop
You may select any VM type available on your cloud. For example:
# 128 Cores and 512 GB of RAM coiled notebook start --vm-type m6i.32xlarge # A single T4 NVIDIA GPU coiled notebook start --vm-type g4dn.xlarge
Coiled can replicate your software and your files to the remote machine, providing a smooth joint development experience.
If you do not select a software environment with the
then Coiled will automatically use Package Synchronization to reproduce your local
versions of software packages on the Jupyter server.
Additionally, if you install mutagen you can use:
coiled notebook start --sync
Then you will get live synchronization between the remote server and your local machine. This allows you to edit files on either machine and have the edits quickly show up on the other.