You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
These command-line scripts support the broader, historical research workspace. They
are not required to read or validate the submitted dissertation. For the preserved
submission workflow, start with make validate.
Public and local utilities
Task
Entry point
Notes
Build the public benchmark dataset
build_labeled_dataset.py
Rebuilds the tracked PhraseBank/FiQA derivative; do not edit the CSV by hand.
Check repository boundaries
check_project.py
Verifies the submission inventory, public-file policy, and documentation links.
Rebuild the GitHub share card
build_social_preview.py
Uses the locked plotting extra; presentation artwork only, with no data or provider calls.
These are historical entry points, not a suggested sequence to rerun. Scripts that
fetch data, score text, or launch remote jobs may use external services, credentials,
licensed inputs, substantial compute, or a recorded authorisation flag. Read the linked
protocol and the script's --help output before running one.
Generated outputs belong in the ignored data, result, report, or experiment-output
locations already defined by the project. Preserve existing manifests and run folders;
the research tools generally create a new artifact or refuse to overwrite one.