Skip to main content
Ctrl+K
pimm pimm

pimm 0.5.2

  • Get started
  • Models
  • Data
  • Training
  • Develop
  • Reference
  • GitHub
  • Hugging Face
  • Get started
  • Models
  • Data
  • Training
  • Develop
  • Reference
  • GitHub
  • Hugging Face

Section Navigation

  • Install
  • First run
  • How pimm works
  • Get started

Get started#

Install pimm, train a tiny model on 100 events, then see how the pieces connect.

  • Install
    • Install with one command
    • Install by hand
    • GPUs
    • Containers
    • Launcher only
    • Settings
    • If the installation fails
  • First run
    • 1. Download the mini dataset
    • 2. Dry-run the job
    • 3. Train
    • 4. Look at what ran
    • 5. Change one thing
    • What happened
  • How pimm works
    • Two kinds of configuration
    • Components are named by registry
    • Events become a packed batch
    • The trainer runs the loop; hooks do the rest
    • Batch sizes are global
    • A run is a directory

previous

pimm

next

Install

© Copyright 2026, DeepLearnPhysics.

Created using Sphinx 8.1.3.

Built with the PyData Sphinx Theme 0.19.0.