Skip to main content

A command-line interface to run ML workflows in the cloud

Project description

A command-line interface to run ML workflows in the cloud


PyPI PyPI PyPI - License Slack

👋 Intro

To run ML workflows, your local machine is often not enough, so you need a way to automate running these workflows using the cloud infrastructure.

Instead of managing infrastructure yourself, writing custom scripts, or using cumbersome MLOps platforms, define your workflows in code and run from command-line.

dstack is an alternative to KubeFlow, SageMaker, Docker, SSH, custom scripts, and many other tools used often for running ML workflows.

Primary features of dstack:

  1. Infrastructure as code: Define workflows and infrastructure requirements declaratively as code. dstack sets up and tears down infrastructure automatically.
  2. GitOps approach: dstack is integrated with Git and tracks code automatically. No need to push local changes before running a workflow.
  3. Artifacts and tags: Artifacts are the first-class citizens. Once the workflow is finished, assign a tag to it and reuse artifacts in other workflows.
  4. Environment setup: No need to build own Docker images or setup CUDA yourself. Just specify Conda requirements and they will be pre-configured.
  5. Interrupted workflows: Artifacts can be stored in real-time, so you can fully-leverage spot/preemptive instances. Resume workflows from where they were interrupted.
  6. Technology-agnostic: No need to use specific APIs in your code. Anything that works locally, can run via dstack.
  7. Dev environments: Workflows may be not only tasks or applications but also dev environments, such as VS Code, JupyterLab, and Jupyter notebooks.
  8. Very easy setup: Just install the dstack CLI and run workflows in your cloud using your local credentials. The state is stored in your cloud storage. No need to set up any complicated software.

📦 Installation

To use dstack, you'll only need the dstack CLI. No other software needs to be installed or deployed.

The CLI will use your local cloud credentials (e.g. the default AWS environment variables or the credentials from ~/.aws/credentials.)

In order to install the CLI, you need to use pip:

pip install dstack

Before you can use dstack, you have to configure the dstack backend:

  • In which S3 bucket to store the state and the artifacts
  • In what region, create cloud instances.

To configure this, run the following command:

dstack config
Configure AWS backend:

AWS profile name (default):
S3 bucket name:
Region name:

The configuration will be stored in ~/.dstack/config.yaml.

That's it. Now you can use dstack on your machine.

✨ Usage

Define workflows

Workflows can be defined in the .dstack/workflows.yaml file within your project.

For every workflow, you can specify the provider, dependencies, commands, what output folders to store as artifacts, and what resources the instance would need (e.g. whether it should be a spot/preemptive instance, how much memory, GPU, etc.)

workflows:
  - name: "train"
    provider: bash
    deps:
      - tag: some_tag
    python: 3.10
    env:
      - PYTHONPATH=src
    commands:
      - pip install requirements.txt
      - python src/train.py
    artifacts: 
      - path: checkpoint
    resources:
      interruptible: true
      gpu: 1

Run workflows

Once you run the workflow, dstack will create the required cloud instance(s) within a minute, and will run your workflow. You'll see the output in real-time as your workflow is running.

$ dstack run train

Provisioning... It may take up to a minute. ✓

To interrupt, press Ctrl+C.

...

Manage tags

Tags help managing data. You can assign tags to finished workflows to reuse their output artifacts in other workflows. Another way to use tags is to upload data to dstack from your local machine and assign n tag to it to use this data in workflows.

Here's how to assign a tag to a finished workflow:

dstack tags add TAG --run-name RUN

Here, TAG is the name of the tag and RUN is the name of the finished workflow run.

If you want to data from your local machine and save it as a tag to use it from other workflows, here's how to do it:

dstack tags add TAG --local-dir LOCAL_DIR

Once a tag is created, you can refer to it from workflows, e.g. from .dstack/workflows.yaml:

deps:
  - tag: some_tag

Manage artifacts

The artifacts command allows you to browse or download the contents of artifacts.

Here's how to browse artifacts:

dstack artifacts list (RUN | :TAG)

Here's how to download artifacts:

dstack artifacts download (RUN | :TAG) [OUTPUT_DIR]

Providers

dstack offers multiple providers that allow running various tasks, applications, and even dev environments.

📘 Docs

More tutorials, examples, and the full CLI reference can be found at docs.dstack.ai.

🛟 Help

If you encounter bugs, please report them directly to the issue tracker.

For questions and support, join the Slack channel.

Licence

Mozilla Public License 2.0

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

dstack-0.0.6.5.tar.gz (49.5 kB view hashes)

Uploaded Source

Built Distribution

dstack-0.0.6.5-py3-none-any.whl (69.7 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page