graphium icon indicating copy to clipboard operation
graphium copied to clipboard

Script for easy download of datasets and splits

Open blazejba opened this issue 2 years ago • 0 comments

Problem: Currently links to download the mixes and splits are hidden in the configuration files, e.g.: https://github.com/datamol-io/graphium/blob/41a1172a16d2da1138c87d4e472d36ba5f8d2e8b/expts/neurips2023_configs/base_config/large.yaml#L71

tdcommons.ai introduces more datasets for single prediction tasks, useful for the evaluation of pre-trained models: https://tdcommons.ai/single_pred_tasks/

Solution: Write a script that allows downloading selected or all datasets and splits easily.

Include:

  • ToyMix,
  • LargeMix,
  • UltraLargeMix,
  • TDC/single_prediction_tasks.

blazejba avatar Sep 08 '23 14:09 blazejba