Allegro Trains Documentation

See our demo Trains Server https://demoapp.trains.allegro.ai/dashboard now.

Trains is Allegro AI’s open source, experimentation and version control solution for data scientists, researchers, and engineers, working individually and collaborating in teams. Trains supports experiment tracking, analysis, reproducibility, comparison, tuning, autoML, storage maintenance, and a variety of additional features. It is a suite of open source Python packages and plugins, including:

  • Trains Python Client Package
    Our Trains Python Client Package allows you to integrate Trains into your experiments with just two lines of code (see Quick Start) and get all of Trains robust automagical logging. Optionally, augment your Python experiment scripts with the powerful features and functionality that our Task, Logger, and Model classes provide.
  • Trains Server
    Our backend infrastructure is Trains Server. It is available to you as either our demo Trains Server (https://demoapp.trains.allegro.ai/dashboard) or your own locally-hosted Trains Server. You can deploy your own Trains Server in a variety of formats, including pre-built Docker images for Linux, Windows 10, Mac OS X, pre-built AWS EC2 AMIs, and Kubernetes standard installations or Kubernetes using Helm. Trains Server includes your own Trains User Interface (Web-App), a RESTful API, and a file server.
  • Trains Agent
    Trains Agent is our DevOps component for experiment execution, resource control, and autoML. You can use Trains Agent with either our demo Trains Server or your own locally-hosted Trains Server (see Installing and Configuring Trains Agent and Trains Agent Reference).
  • Trains Plugins
    Trains provides plugins to assist you and improve your productivity, including the Trains Jupyter Plugin adding Git support, and the Trains PyCharm Plugin allowing you to easily synchronize local repository information to a remote debugger machine.