This is a reupload of DS-1000. The metadata dictionary has been extracted into columns and the categorical variables are now ClassLabel types, and the dataset is natively a parquet. The features are as follows:
Column
Type
problem_id
Value(dtype='int64', id=None)
prompt
Value(dtype='string', id=None)
reference_code
Value(dtype='string', id=None)
code_context
Value(dtype='string', id=None)
library_problem_id
Value(dtype='int64', id=None)
library… See the full description on the dataset page:
https://huggingface.co/datasets/claudios/ds1000.
