Hugging Face's logo

Datasets:
HuggingFaceFW
/
fineweb

Tasks:
Text Generation
Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
Size:
10B - 100B
ArXiv:
DOI:
Libraries:
Datasets
Dask
License:
Dataset card Viewer Files Community
50
New discussion

350BT sample is much smaller than advertized
#53 opened 23 days ago by DavidNemeskey

Updates
#52 opened 25 days ago by oscarwang2

Exact copy of this dataset on HuggingFace yields "This dataset has 218 files that have been marked as unsafe."

1
#50 opened about 1 month ago by egor-pakhomov

Simple exact deduplication removes 2/3 of data.

3
#49 opened about 1 month ago by egor-pakhomov

Individual files are too large to be used in some Parquet readers
#48 opened about 1 month ago by Vectorrent

Truly the finest of the web
#47 opened about 2 months ago by takeraparterer

requests.exceptions.HTTPError: 503 Server Error: Service Unavailable for url: https://huggingface.co/api/datasets/HuggingFaceFW/fineweb-edu/paths-info/5b89d1ea9319fe101b3cbdacd89a903aca1d6052

2
#46 opened about 2 months ago by Rainyyyday

Dataset Viewer issue: JobManagerCrashedError

1
#45 opened 2 months ago by Enderchef

MMLU evaluation setting
#44 opened 2 months ago by jordane95

Fineweb Download Size Discrepency
#43 opened 3 months ago by msmmpts

Is there an official test set for benchmarking objectively?

1
#42 opened 3 months ago by SophieOstmeier

Any plans to release warc content after the language filtering steps?

1
#41 opened 3 months ago by Splend1dchan

Casting Issue?

4
#40 opened 3 months ago by FelixLabelle

Fineweb train configuration

1
#39 opened 3 months ago by nezhazheng

Reproducibility of the work for other languages

2
#38 opened 3 months ago by camillop

why do you apply "All filters except the (very destructive) terminal_punct"

2
#36 opened 3 months ago by bpwl0121

How to compute the aggerate score?

1
#35 opened 3 months ago by mornmirror

Language subset

1
#33 opened 3 months ago by talmor

Dedup

1
#32 opened 3 months ago by shawnkx

Regarding to the newly updated indexes(writen as deduplication issues)

5
#29 opened 4 months ago by kimcando

"Reproducing GPT-2 (124M) in llm.c in 90 minutes for $20" using fineweb by Karpathy
#28 opened 4 months ago by clem

Unsafe files
#25 opened 5 months ago by alielfilali01

tiny-fineweb

2
#19 opened 5 months ago by 3thn

Training configs for data ablation study

1
#14 opened 5 months ago by jimmyhbx

Reprocessing for a new language

13
#12 opened 5 months ago by pere

Are copyrighted works included in this dataset?

4
#9 opened 5 months ago by umm-maybe

Any plan to train models on larger subset of dataset?

1
#8 opened 5 months ago by mrfakename

Torrent?

2
#4 opened 5 months ago by emilss