Hugging Face's logo

Datasets:
PatrickHaller
/
pile-100M-words

Modalities:
Text
Formats:
parquet
Languages:
English
Size:
100K - 1M
Libraries:
Datasets
Dask
Dataset card Viewer Files Community
3
PatrickHaller commited on
Commit
ad2fcbe
·
verified ·
1 Parent(s): f457ef1

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -2
README.md CHANGED
@@ -28,7 +28,6 @@ configs:
28
  # Description
29
 
30
  This dataset is a sampled subset of the [Pile](https://huggingface.co/datasets/EleutherAI/pile) dataset.
31
- We used [DSIR](https://github.com/p-lambda/dsir) a data selection tool with importance resampling for subsampling.
32
 
33
  The subset sample distribution is:
34
 
@@ -63,7 +62,7 @@ The dataset contains ~100M words of text. This can be checked with:
63
  ```python
64
  from datasets import load_dataset
65
 
66
- ds = load_dataset("PatrickHaller/dsir-pile-100M-words")
67
 
68
  count = 0
69
  for row in ds["train"]:
 
28
  # Description
29
 
30
  This dataset is a sampled subset of the [Pile](https://huggingface.co/datasets/EleutherAI/pile) dataset.
 
31
 
32
  The subset sample distribution is:
33
 
 
62
  ```python
63
  from datasets import load_dataset
64
 
65
+ ds = load_dataset("PatrickHaller/pile-100M-words")
66
 
67
  count = 0
68
  for row in ds["train"]: