Bundled Models Index

Unlike the other directories in the ‘packages’ directory of PyEarthTools, the “bundled_models” directory does not itself contain a “bundled models” Python package. Rather, it contains multiple model packages in separate directories. Each of these bundled models is a Python package. As such, “bundled_models” is not itself installable. This page will provide an index table for each bundled model.

At the current time, the following bundled models are available:

• FourCastNeXt by Guo et al. (2024).

• LUCIE by Guan et al. (2025).

Bundled models also have configuration files in addition to the the Python code. Each yaml file is also included in the table for the bundled model.

The API docs for each bundled model will also be presented together in the Bundled Models API Docs.

FourCastNeXt Bundled Model

Module	Purpose	API Docs
fourcastnext	PyTorch Lightning API	- lightning_model.FourCastNextML
	PyEarthTools Registration Interface	- registered_model
	Crop ERA5 grid to required spacing	- CropToRectangle
	Crop ERA5 low res grid to required spacing	- CropToRectangleSmall
fourcastnext.architecture	Multilayer Perceptron	- Mlp
	2D AFNO Network	- AFNO2D
		- Block
		- AFNONet
	Patching and embedding	- PatchEmbed
Training Directory	Configuration for different experiments
training/configs	FourCastNeXt default configuration	config.yaml
	Worker and batch size for data preprocessing	data/module/default.yaml
	Default model data split	data/splits/default.yaml
	Data patch size and number of channels	data/example.yaml
	Train on a reduced data set	splits/short_training.yaml
	PyTorch model initialisation parameters	model/default.yaml
	Training strategy configuration	trainer/default.yaml
training/limited_variables_early_stopping	FourCastNeXt full-size reduced-training configuration	limited_vars_early_stop.yaml
	FourCastNeXt low resolution configuration	lowres.yaml
	Worker and batch size for training	module/default.yaml
	Full-length training period	splits/default.yaml
	Reduced-length training period	splits/short_training_splits.yaml
	Reduced-length training period	splits/short_training_splits.yaml
	Full-res data and channels	data/example.yaml
	Low-res data and channels	data/lowres.yaml
	PyTorch model initialisation parameters	model/default.yaml
	Training strategy for full convergence	trainer/default.yaml
	Train a reduced number of epocs	trainer/few_epochs.yaml
Pipelines Directory	Define data normalisation pipeline
pipelines	Full-resolution pipeline	early_stopping.pipe
	Low-resolution pipeline	low_res_demo_subset.pipe
	General random data for testing	example.pipe