Dynamic World Model Runner

August 8, 2025 ยท View on GitHub

DOI

Model files and example notebook for Dynamic World, see https://doi.org/10.1038/s41597-022-01307-4.

This is not an officially supported Google product.

  • TensorFlow SavedModels for the forward and backward path can be found in ./model/forward and ./model/backward respectively.
  • ./single_image_runner.ipynb gives a step-by-step guide to making predictions with the Dynamic World models.

See the image below for the model block diagram. For Dynamic World, m = 1.5 and b = 2.

Dynamic World neural network architecture