Why battery-hub
Battery AI results are hard to compare. Datasets use different formats, tasks are defined differently in every paper, and pretrained models are rarely shared. Inspired by Hugging Face, this hub provides the missing layer: standard tasks, models with cards and documented datasets.
How it works
Task registryEvery task, dataset and model is a YAML file in a public git repository. Adding one is a pull request, reviewed like code.
Storage on Hugging FaceData files and model weights live in the battery-hub organization on Hugging Face. This site indexes them.
Battery Data FormatExperimental cycling data uses the Battery Data Format (BDF) from the LF Energy Battery Data Alliance. Synthetic curves and labels use Parquet with BDF-style column names.
EvaluationLeaderboard scores come from a shared evaluation harness run against fixed splits, so numbers are comparable across models.