ToolJovia
Data

Grouped Dataset Partition Bundle

Create reproducible group-atomic CSV partitions with explicit group weights and a content-hash membership manifest.

✓ Browser-local processing
Processed in this tab. Up to 100,000 characters unless this tool states a smaller limit.

Loading tool…

How to use Grouped Dataset Partition Bundle

Keep related records together when preparing separate local datasets. Exact group tuples are assigned as whole units, and the manifest distinguishes group quotas from row counts instead of implying balanced labels, independent subjects or a leakage-free experiment.

  1. Enter input, atomic group key columns json, portable partition names and integer weights json, reproducibility seed.
  2. Choose the settings for grouped dataset partition bundle, then run the tool.
  3. Review the result and any limitations before copying or downloading it.

Good to know

  • 2,000 source rows and 200,000 CSV characters (50,000 per cell, 100 columns), 2–5 partitions with exact original integer weights 1–1,000, 100 seed characters and 1 MiB combined generated content before stored ZIP allocation.
  • Group counts follow largest remainders, not row counts or stratified label balance. Grouping correctness is supplied, not inferred. Same keys/seed/weights reproduce membership; adding/removing groups can change existing membership. Hashes do not anonymize the intentionally included source data.
  • Processing is local. Output is limited to 1,000,000 text characters unless a stricter file limit is stated.

Common questions

Does a group-safe split prove there is no train/test leakage?

No. It only keeps identical supplied key tuples together. Related entities can still appear under different keys, and time, feature, label or preprocessing leakage requires separate domain-aware checks.

Is this input sent to a server?

No. The application processes this input inside your browser tab and does not save it to persistent storage.

Keep going with related tools

All data →
Data

CSV Stratified Sampler

Sample up to a declared number of CSV records from each exact group without replacement.

Data

CSV Foreign Key Checker

Check child records against unique parent composite keys with explicit blank-key handling.

Data

Tabular Fixture Dataset Builder

Export deterministic related CSV tables and an exact-string JSON bundle from bounded original recipes.