Datasetpublic

Aipresso/medium_512_1k_tokens_prompts

Medium 512-1K Tokens Prompts Dataset Created by Aipresso LIMITED, London, UK ⚠️ By using this dataset you agree to our Terms of Use. Overview 703 high-quality English prompts whose length lies between 512 and 1 000 tokens.Every prompt has been de-duplicated, cleaned and token-counted with the GPT-2 tokenizer. Statistics Rows Token range File size Format 703 512 – 1 000 2.9 MB CSV Use-cases Medium-context… See the full description on the dataset page: https://huggingface.co/datasets/Aipresso/medium_512_1k_tokens_prompts.

sourceHugging Faceotherupdated 11mo agoView on Hugging Face

0likes33downloads

Dataset Card

Medium 512-1K Tokens Prompts Dataset

Created by Aipresso LIMITED, London, UK

⚠️ By using this dataset you agree to our [Terms of Use](TERMS_OF_USE.md).

Overview

703 high-quality English prompts whose length lies between 512 and 1 000 tokens. Every prompt has been de-duplicated, cleaned and token-counted with the GPT-2 tokenizer.

Statistics

Rows	Token range	File size	Format
703	512 – 1 000	2.9 MB	CSV

Use-cases

—Medium-context language-model fine-tuning
—Research on optimal context-length trade-offs
—Enterprise systems that need more than “short” prompts but less than full documents

File layout

data/ └── medium5121ktokensprompts.csv Copy

Columns

—Prompt (str): the cleaned prompt text
—Token_count (int64): exact GPT-2 token count (512-1 000)

Quick start

python

from datasets import load_dataset
ds = load_dataset("Aipresso/medium_512_1k_tokens_prompts", split="train")
print(ds[0])
Citation
bibtex
Copy
@dataset{aipresso_medium_prompts_2024,
  title={Medium 512-1K Tokens Prompts Dataset},
  author={Aipresso LIMITED},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/Aipresso/medium_512_1k_tokens_prompts}
}
Related datasets
MEGA Cleaned Prompts – 2.7 M rows
10K Cleaned Prompts – mixed lengths
<512 Tokens – short prompts
>1K Tokens – long prompts
Maintained by Aipresso LIMITED, London, UK.