TorchX is a universal job launcher for PyTorch applications. TorchX is designed to have fast iteration time for training/research and support for E2E production ML pipelines when you're ready.

TorchX currently supports:

Kubernetes (EKS, GKE, AKS, etc)
Slurm
Docker
Local

Need a scheduler not listed? Let us know!

Quickstart

See the quickstart guide.

Documentation

Requirements

torchx:

python3 (3.8+)
PyTorch
optional: Docker (needed for docker based schedulers)

Certain schedulers may require scheduler specific requirements. See installation for info.

Installation

Stable

# install torchx sdk and CLI -- minimum dependencies
pip install torchx
# install with all dependencies
pip install "torchx[dev]"
# install torchx Kubernetes / Volcano support
pip install "torchx[kubernetes]"
# install torchx GCP Batch support
pip install "torchx[gcp_batch]"

Nightly

# install torchx sdk and CLI
pip install torchx-nightly[dev]

Source (Development)

# clone the repo
git clone https://github.com/meta-pytorch/torchx.git
cd torchx
# install with uv (recommended for development)
uv sync --extra dev
# or with pip
pip install -e ".[dev]"

Docker

TorchX provides a docker container image at ghcr.io/pytorch/torchx for use as part of a TorchX role.

Contributing

We welcome PRs! See the CONTRIBUTING file.

License

TorchX is BSD licensed, as found in the LICENSE file.