Skip to content

Installation

Choose how you want to run the TServe server. Docker is the fastest and recommended way: each image already contains the dependencies for its model family, and separate CPU and GPU tags remove any torch setup work.

Docker

Install Docker Desktop on macOS or Windows, or Docker Engine on Linux.

Pull the hub image, which supports Chronos Bolt/T5, TTM, and TimesFM 2.x:

docker pull sktime/tserve:hub

NVIDIA hosts need the NVIDIA Container Toolkit and --gpus all at runtime:

docker pull sktime/tserve:hub-gpu

Other model families use different image tags. Choose the model first, then use its tag from the model catalog. Every CPU tag has a -gpu variant. For Hugging Face tokens, cache volumes, and all Docker options, see Docker.

UV / Pip

TServe requires Python 3.12 or newer. Install the server extra and the extra for the model family you need. The examples below install the hub family. A plain install pulls the CUDA build of torch (MPS on macOS); the CPU tabs skip that download on a machine without a GPU.

Install uv, create a virtual environment, and install TServe:

uv venv

Then install TServe:

uv pip install "tserve[server,hub]"

uv pip install torch --index-url https://download.pytorch.org/whl/cpu

uv pip install "tserve[server,hub]"

Create a virtual environment, activate it, and install TServe:

python -m venv .venv
source .venv/bin/activate

py -m venv .venv
.venv\Scripts\Activate.ps1

Then install TServe:

python -m pip install "tserve[server,hub]"

python -m pip install torch --index-url https://download.pytorch.org/whl/cpu

python -m pip install "tserve[server,hub]"

The server extra alone supports the naive test baseline. Replace hub with another family extra, or use full for every family. The same CPU-first order is documented as CPU-only install.

From source

Use a source install when developing TServe or testing unreleased changes. It is also the only path where the gpu extra selects the torch index, because that choice lives in the repository's uv lockfile. The From source guide covers cloning the repository, editable installs, dependency extras, and GPU setup.

Next

Continue to the Quick start to launch the server and send your first prediction.