torchx
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
What is torchx?
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.
Overview of torchx
What it is
TorchX is a universal job launcher for PyTorch applications. It is designed to have fast iteration time for training/research and support for E2E production ML pipelines when you are ready. The tool currently supports Kubernetes (EKS, GKE, AKS, etc), Slurm, Docker, and local execution.
Key features
- Supports multiple schedulers: Kubernetes, Slurm, Docker, and local.
- Designed for rapid iteration during training and research.
- Provides a CLI and SDK for launching jobs.
- Can be extended to support E2E production ML pipelines.
- Offers nightly and stable installation channels.
- Provides a Docker container image at ghcr.io/pytorch/torchx.
Use cases
TorchX is used for launching PyTorch training jobs across different environments, from local development to production clusters. It is suitable for researchers who need fast iteration cycles and for engineers transitioning workloads to production on Kubernetes, Slurm, or cloud batch services.
Getting started
Installation instructions from the README:Stable:bashpip install torchxpip install torchx[dev]Nightly:bashpip install torchx-nightly[dev]For Kubernetes, GCP Batch, or source installation, see the repository README.
License and pricing
TorchX is BSD licensed, as found in the LICENSE file. It is an open source project with no stated pricing.
Based on the project's public README and repository metadata.