Datasets:

pietrolesci
/

pile-deduped-pythia-preshuffled

Tasks:

Text Generation

Modalities:

Formats:

Languages:

Size:

Libraries:

Dataset card Data Studio Files Files and versions

pietrolesci commited on Mar 25

Commit

1b0c508

·

verified ·

1 Parent(s): 6f94734

Update README.md

Files changed (1) hide show

README.md +9 -0

README.md CHANGED Viewed

@@ -5,6 +5,15 @@ language:
 - en
 size_categories:
 - 100B<n<1T
 ---
 This dataset contains the fully prepared data, which has been tokenized and pre-shuffled, used to train the Pythia (deduplicated) models.

 - en
 size_categories:
 - 100B<n<1T
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: data/*.parquet
+  - config_name: detokenized
+    data_files:
+      - split: train
+        path: detokenized/*.parquet
 ---
 This dataset contains the fully prepared data, which has been tokenized and pre-shuffled, used to train the Pythia (deduplicated) models.