Skip to content

Provision Kaggle credentials for MLE-Bench dataset preparation #131

Description

@micahstubbs

Board issue: sia_rust-e2j.6

Finding

The local environment does not define KAGGLE_USERNAME and KAGGLE_KEY. That blocks MLE-Bench dataset preparation when using the Kaggle API path documented for this project.

The project can also use ~/.kaggle/kaggle.json; this audit did not find env vars in the local runtime. The issue should confirm whether a local Kaggle token file exists, and if not, provision credentials.

Code path

  • docs/walkthrough.md:111 documents prepare_mlebench_dataset.py for MLE-Bench tasks.
  • docs/walkthrough.md:116 requires KAGGLE_USERNAME and KAGGLE_KEY for Kaggle API downloads.
  • docs/walkthrough.md:120 documents ~/.kaggle/kaggle.json as the alternate credential location.
  • sia/prepare_mlebench_dataset.py:31 invokes mlebench prepare, which performs the dataset download.

Acceptance criteria

  • Confirm whether ~/.kaggle/kaggle.json is present and usable; if not, add Kaggle credentials to secure local/team secret storage.
  • Export KAGGLE_USERNAME and KAGGLE_KEY or install the Kaggle token file with correct permissions.
  • Accept any required Kaggle competition rules before running mlebench prepare.
  • Verify one python -m sia.prepare_mlebench_dataset ... --skip-gemini preparation path and attach only non-secret output.

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions