Skip to content

Comments

docs: clarify dataset split instructions (train/val/test vs trainval)#427

Open
dhruvildarji wants to merge 1 commit intomotional:masterfrom
dhruvildarji:docs/clarify-dataset-split-instructions
Open

docs: clarify dataset split instructions (train/val/test vs trainval)#427
dhruvildarji wants to merge 1 commit intomotional:masterfrom
dhruvildarji:docs/clarify-dataset-split-instructions

Conversation

@dhruvildarji
Copy link

Summary

  • Add a new "Dataset splits" section to docs/dataset_setup.md explaining that all .db files from the separately downloaded train, val, and test archives should be placed into the single trainval/ directory
  • Clarify that the devkit handles logical train/val/test splitting internally via the log splitter configuration (nuplan/planning/script/config/common/splitter/nuplan.yaml)
  • Annotate the directory tree to make the trainval folder's purpose explicit

Fixes #417 (also addresses #342)

Motivation

The download page provides separate archives for train, val, and test splits, but the existing documentation only showed a trainval directory without explaining the relationship. This caused confusion for new users who did not know whether to create separate directories or merge the files.

Test plan

  • Verify the rendered documentation reads clearly
  • Confirm the directory tree annotation displays correctly in Markdown

The download page provides separate train, val, and test archives, but
the documentation did not explain that all .db files should be placed
into a single trainval/ directory. This caused confusion for new users
who expected separate directories for each split.

Add a new section explaining the relationship between downloaded splits
and the trainval directory, and annotate the directory tree accordingly.

Fixes motional#417

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

Clarify dataset split instructions in documentation (train, val, test vs trainval)

1 participant