spott/dstack

dstack is an open-source engine for running GPU workloads. It simplifies development, training, and deployment of gen AI models on any cloud. Discord: https://discord.gg/u8SmfwPpMd

Orchestrate GPU workloads effortlessly on any cloud

Docs • Examples • Discord

dstack is an open-source engine for running GPU workloads on any cloud. It works with a wide range of cloud GPU providers (AWS, GCP, Azure, Lambda, TensorDock, Vast.ai, etc.) as well as on-premises servers.

Latest news ✨

• [2024/03] dstack 0.16.1: Improvements to dstack pool and bug-fixes (Release)
• [2024/02] dstack 0.16.0: Pools (Release)
• [2024/02] dstack 0.15.1: Kubernetes integration (Release)
• [2024/01] dstack 0.15.0: Resources, authentication, and more (Release)
• [2024/01] dstack 0.14.0: OpenAI-compatible endpoints (Release)
• [2023/12] dstack 0.13.0: Disk size, CUDA 12.1, Mixtral, and more (Release)

Installation

Before using dstack through CLI or API, set up a dstack server.

Install the server

The easiest way to install the server, is via pip:

pip install "dstack[all]" -U

Configure backends

If you have default AWS, GCP, or Azure credentials on your machine, the dstack server will pick them up automatically.

Otherwise, you need to manually specify the cloud credentials in ~/.dstack/server/config.yml.

For further details on setting up the server, refer to installation.

Start the server

To start the server, use the dstack server command:

$ dstack server

Applying ~/.dstack/server/config.yml...

The admin token is "bbae0f28-d3dd-4820-bf61-8f4bb40815da"
The server is running at http://127.0.0.1:3000/

Note It's also possible to run the server via Docker.

CLI & API

Once the server is up, you can use either dstack's CLI or API to run workloads. Below is a live demo of how it works with the CLI.

Dev environments

You specify the required environment and resources, then run it. dstack provisions the dev environment in the cloud and enables access via your desktop IDE.

Tasks

Tasks allow for convenient scheduling of any kind of batch jobs, such as training, fine-tuning, or data processing, as well as running web applications.

Specify the environment and resources, then run it. dstack executes the task in the cloud, enabling port forwarding to your local machine for convenient access.

Services

Services make it very easy to deploy any kind of model or web application as public endpoints.

Use any serving frameworks and specify required resources. dstack deploys it in the configured backend, handles authentication, and provides an OpenAI-compatible interface if needed.

Pools

Pools simplify managing the lifecycle of cloud instances and enable their efficient reuse across runs.

You can have instances provisioned in the cloud automatically, or add them manually, configuring the required resources, idle duration, etc.

Examples

Here are some featured examples:

• TGI
• vLLM
• Ollama
• SDXL
• QLoRA

Browse dstack.ai/examples for more examples.

More information

For additional information and examples, see the following links:

• Docs
• Discord

Licence

Mozilla Public License 2.0