L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Données et recherche · SKILL.md

dataset-contribution

dataset-contribution

Routes OGB dataset-export workflows for DatasetSaver and OGB-compatible releases.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: dataset-contribution description: "Routes OGB dataset-export workflows for DatasetSaver and OGB-compatible releases." metadata: disco-role: operating disable-model-invocation: true license: MIT

Dataset Contribution

Use this subskill when the task is about packaging a new OGB-compatible dataset with DatasetSaver, generating a meta_dict.pt, or validating the layout of a submission archive.

Use this subskill when

  • The task names DatasetSaver or save_graph_list / save_target_labels / save_split / copy_mapping_dir / get_meta_dict.
  • The task asks how to prepare meta_dict.pt, mapping/, or a release zip for an external dataset contribution.
  • The task needs a tiny end-to-end smoke of the export pipeline.

First decisions

  1. Read references/workflows.md for the step order and family constraints.
  2. Read references/data-formats.md for graph, split, label, and mapping directory shapes.
  3. Read references/api-reference.md for the method order and parameter expectations.
  4. Read references/troubleshooting.md when a shape, mapping, or zip step fails.
  5. If you want a safe end-to-end smoke, run scripts/datasetsaver-tiny-smoke.py.

Main workflow

  • Decide whether the dataset is graph-, node-, or link-oriented.
  • Build the graph list in the format OGB expects.
  • Save labels and splits in the same order that the export helper expects.
  • Copy the mapping directory only after the raw and split files are present.
  • Generate the meta_dict.pt and zip the prepared directory.
  • Reload the tiny smoke dataset to confirm the archive structure still works.

Common routing choices

  • Use this subskill for new OGB-compatible releases.
  • Use the graph/node/link subskills when the user only needs to consume an existing OGB dataset, not package a new one.

What not to do here

  • Do not tell the user to hand-edit the packaged release after the zip step.
  • Do not assume heterogeneous ogbg export is implemented.
  • Do not depend on the original checkout for final runtime guidance.

Related references

Skills enregistrées