Datasets:
Tasks:
Text Classification
Formats:
parquet
Sub-tasks:
language-identification
Size:
1M - 10M
License:
Commit History
Yesterday
-
Dataset card: how it relates to the models' build
Verified signature8bf5845 -
Dataset card: one paragraph per line, the released models; Unix line endings
Verified signature8021a70 -
Language identification: 1641 labels, 2390082 passages
Verified signature614b76c -
initial commit
Verified signature0e7176e