motional
Clarify dataset split instructions in documentation (train, val, test vs trainval)
Description
I followed the instructions in the dataset setup documentation, but I found some parts to be unclear.
Specifically, the documentation mentions trainval as a dataset split, while the actual data is downloaded as train, val, and test. It is not clear whether we need to manually merge train and val into trainval, or if test should be added separately.
Related Problem
The ambiguous instructions about dataset splits (train/val/test vs trainval) in the documentation led me to confusion and required additional verification steps.
Describe alternatives you've considered
I tried checking the dataset directory structure and related code to infer how the splits should be organized, but an explicit explanation in the documentation would help avoid confusion.
Additional context
Since this document is often the first reference for users unfamiliar with nuplan-devkit, improving clarity in this section would greatly help new users avoid common misunderstandings.
A brief explanation about how trainval relates to train and val, and how to handle test, would be very beneficial.
This issue is also discussed in Issue #342.
Additionally, as noted in this comment, there appears to be an initial path configuration issue. The documentation should clearly state whether splits need to be added or modified to avoid confusion.