update dataset upload method to include users with dashes in their usernames

#3
by not-lain - opened
Hugging Face Agents Course org

currently you can't create splits for users with dashes in their names, some usernames that do not work with the previous implementation are not-lain, julien-c, etc....
by switching to this implementation we upload a file called [username].parquet for each username, I also made it compatible with the dataset viewer (see example dataset and example space)

cc @burtenshaw for review

Hugging Face Agents Course org
edited 5 days ago

some observations :

  • create an empty dataset in this organization before merging according to the name i chose agents-course/students-data
  • for the future we should check for the max value of score between the current score and the previous score
  • if you choose to change the column names, do it from the beginning since if you change them later the dataset viewer will be broken (schema is defined in the data_to_parquet.py)
Hugging Face Agents Course org

closing this as it has been fixed in 2606c173391898414a06f161496e19835ff9ffae

not-lain changed pull request status to closed

Sign up or log in to comment