meta-pytorch/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.

What it solves

TorchX simplifies the process of launching and managing PyTorch applications across different computing environments. It removes the need for users to write complex, environment-specific configuration files for various cluster managers or cloud providers, providing a universal interface for job submission.

How it works

TorchX acts as a universal job launcher. It abstracts the underlying infrastructure, allowing users to launch PyTorch jobs on a variety of supported schedulers, including Kubernetes (EKS, GKE, AKS), Slurm, AWS Batch, Docker, and local machines. This allows for a fast iteration cycle during research and a seamless transition to production ML pipelines.

Who it’s for

This tool is designed for PyTorch developers and ML researchers who need to run their code on a variety of different compute resources without having to learn the specific configuration requirements of each scheduler.

Highlights

  • Universal job launcher for PyTorch applications.
  • Supports multiple schedulers: Kubernetes, Slurm, AWS Batch, Docker, and local.
  • Designed for both fast research iteration and end-to-end production ML pipelines.
  • Provides a CLI and SDK for job submission.