Skip to content

Dataset · HuggingFaceFW

FineWeb-Edu Synthetic Phrases

HuggingFaceFW/finephrase

English synthetic text data generated by SmolLM2 from the FineWeb-Edu corpus.

What's inside

The dataset contains English synthetic texts created by DataTrove using the SmolLM2-1.7B-Instruct model. The source is the sample-350BT configuration of the fineweb-edu set, train split. Generation used temperature=1.0, top_p=1.0, top_k=50, and max_tokens=2048; the card also lists the faq prompt family and the text column.

What you can do with it

  • Fine-tune a small English LLM on synthetic texts
  • Test generation with SmolLM2-1.7B-Instruct
  • Investigate the impact of synthetic data on text-generation
Tasks
Text generation
Languages
English
Size
1B–10B rows
Downloads this month
171.5K
Likes
165

How to load it

from datasets import load_dataset

ds = load_dataset("HuggingFaceFW/finephrase")