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 to write environment-specific configuration files for every different scheduler, allowing researchers and engineers to move from local development to large-scale production pipelines without changing their code.

How it works

TorchX acts as a universal job launcher. It provides a unified interface to submit jobs to various schedulers and infrastructure providers, including Kubernetes (EKS, GKE, AKS), Slurm, Docker, and local machines.

Who it’s for

It is designed for PyTorch users who need to iterate quickly during the research phase and then scale their workloads to production ML pipelines.

Highlights

  • Universal job launcher for PyTorch applications.
  • Support for multiple schedulers including Kubernetes, Slurm, and Docker.
  • Fast iteration time for training and research.
  • Support for end-to-end production ML pipelines.

Related

  • Project
    meta-pytorch/torchcomms

    An experimental communications API for PyTorch that provides a high-level collectives interface and support for multiple hardware backends like NCCL, RCCL, and XCCL.

  • Project
    datawhalechina/torch-rechub

    Torch-RecHub is a PyTorch framework for building recommender systems with 30+ models, modular design, ONNX export, and support for multiple hardware backends.

  • Project
    duoan/TorchCode

    An interactive coding platform for practicing PyTorch implementations of operators and architectures from scratch to prepare for ML engineering interviews.

  • Project
    alibaba/TorchEasyRec

    A PyTorch-based recommendation system framework that provides pre-implemented state-of-the-art models and scalable infrastructure for production-ready deep learning recommendations.

  • Project
    MooreThreads/torch_musa

    torch_musa is a PyTorch extension that adds a **MUSA** backend for Moore Threads GPUs. It mirrors the standard PyTorch API, so you only replace `cuda` with `musa`. The package includes tools for building C/C++ extensions, distributed training via the **mccl** backend, and Docker images for easy setup. It also provides MUSA‑compatible builds of torchvision, torchaudio, and a curated list of other popular PyTorch libraries. Installation can be done from pre‑built wheels or by building from source after installing the MUSA SDK. The project is BSD‑licensed.