Prepare. Train. Finetune. Reproduce.

Define reproducible deep learning workflows. Run them via the CLI. Use any infrastructure. Browse logs and artifacts in real-time.

Sign up to join the preview program
Dstack website hero
Device frame


For each workflow, specify a Docker image, commands, other workflows it depends on, and output artifacts.

Dstack website zoom workflows


Each workflow may have variables and their default values. Any of the variables can be overridden when you run the workflow via the CLI.

Dstack website zoom vars

Self-hosted runners

Register your local machine or any remote servers as self-hosted runners. To do that, you only need to run one line of bash code.

After you set up self-hosted runners, submit any workflow to run there from your laptop using the CLI.

Dstack website zoom self hosted

Spot instances

Aulternatively to self-hosted runners, you can authorize dstack to set up spot instances in your own cloud account.

If any of the spot instances terminate before finishing the workflow, dstack will re-assign it to a new instance without losing the checkpoints.

Dstack website zoom autoscale 1


Run workflows via the CLI. Browse real-time logs and artifacts. Stop and resume runs.

Mark sucessful runs with tags for later reuse.

Dstack website zoom run


Got a question?

Drop in to Slack to ask our friendly experts, or search our docs yourself.

Sign up to join the preview program

Runs on Unicorn Platform