ocp / data /tutorials /README.md
introvoyz041's picture
Migrated from GitHub
b78a213 verified

Tutorials

As part of the NeurIPS 2021 Climate Change AI Workshop, a comprehensive, interactive Google-Colab tutorial notebook can be found here. This notebook is designed for those new to OC20 and interested in how to get started. Topics include:

  • Background
  • Software Requirements
  • Dataset overview & Visualization
  • OCP Tasks - Train, Validate, Predict
    • IS2RE
    • S2EF
    • IS2RS
  • OCP Calculator
  • Model Development

Additionally, we provide several Jupyter notebooks:

  • Data preprocessing - preprocessing raw ASE atoms object to OCP graph Data objects.
  • LMDB dataset creation - creating your own OCP-comatible LMDB datasets from ASE-compatible Atoms objects.
  • Data visualization - understanding the raw data and its contents. (same contents found in the above google colab notebook)
  • S2EF training example - training a SchNet S2EF model, loading a trained model, and making predictions. (same contents found in the above google colab notebook).