gpjt/fineweb-gpt2-tokens
FineWeb GPT-2 Tokens This is all of the 10B sample of FineWeb, split into 99% training data, 1% validation data, each converted to a single uint16 tensor of GPT-2 tokens. The text for each record in the original dataset was tokenized using tiktoken, then they were all concatenated together, with <|endoftext|> tokens separating them. Code: Script to download the original dataset Script to do the tokenization Based on This dataset is derived from the original… See the full description on the dataset page: https://huggingface.co/datasets/gpjt/fineweb-gpt2-tokens.
| Author / organization | gpjt |
|---|---|
| License (as declared) | odc-by |
| Downloads | 125 (as reported by Hugging Face) |
| Likes | 0 |
| Created | 4 December 2025 |
| Last updated at source | 5 December 2025 |
Popularity, last 2 days
125 now (-1 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
- gpjt/8xa100m40-lore-6-new-seed-baseline-disabledModel · used to train
- gpjt/8xa100m40-lore-10-both-corrected-smart-initModel · used to train
- gpjt/8xa100m40-lore-2-both-ordinary-initModel · used to train
- gpjt/8xa100m40-lore-11-output-only-corrected-smart-initModel · used to train
- gpjt/8xa100m40-lore-8-new-seed-both-smart-initModel · used to train
- gpjt/8xa100m40-lore-5-output-onlyModel · used to train
- gpjt/8xa100m40-lore-9-new-seed-output-onlyModel · used to train
- gpjt/8xa100m40-lore-1-baseline-disabledModel · used to train
- gpjt/8xa100m40-lore-3-both-smart-initModel · used to train
- gpjt/8xa100m40-lore-4-input-onlyModel · used to train
- gpjt/8xa100m40-lore-7-new-seed-input-onlyModel · used to train
- gpjt/8xa100m40-baselineModel · used to train
Recent changes
- 16 Oct 2026 Description: FineWeb GPT-2 Tokens This is all of the 10B sample of FineWeb, split into 99% training data, 1% validation data, each converted to a single uint16 tensor of GPT-2 tokens. The text for each record in the original dataset was tokenized using → FineWeb GPT-2 Tokens This is all of the 10B sample of FineWeb, split into 99% training data, 1% validation data, each converted to a single uint16 tensor of GPT-2 tokens. The text for each record in the original dataset was tokenized using
Changes detected by Silytics between refreshes of the source.
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.