How-to Tutorials

• How to prepare training dataset
  • What does NequIP behind the scene
  • Key concepts
    • fixed_fields
    • yaml interface
    • Python interface
  • Prepare dataset and specify in yaml config
    • ASE format
    • NPZ format
    • Note on key mapping
  • Advanced options
    • skip frames during data processing
    • register user-defined graph, node, edge fields
• How to migrate to newer NequIP versions
  • Upgrade NequIP
    • 1. Record the old version
    • 2. Update your main nequip repo
  • Obtain the state_dict from the old ver
    • 1. Back up the old version
    • 2. Save the state_dict from the old verion
  • Load the state_dict in the new version
  • Validate the result using nequip-evaluate
    • Old model
    • New model