b-remy/

neurips-wl-challenge-split

This dataset is a HF mirror of the official challenge training data for this challenge: https://www.codabench.org/competitions/8934/ This repo contains a preprocessed train/validation split. The original data has shape ncosmo, np, .... This dataset is splitted along the np dimension with a fraction of 0.8, and reordered as ncosmo, np, ... -> np, ncosmo, ...

To get started: import datasets

dset_train =… See the full description on the dataset page: https://huggingface.co/datasets/b-remy/neurips-wl-challenge-split.