Skip to content

Dataset · yahma

Cleaned Alpaca dataset

yahma/alpaca-cleaned

A corrected version of Stanford Alpaca for instruction fine-tuning in English.

What's inside

A cleaned version of the original Stanford Alpaca containing English instruction-response text pairs. The authors fixed hallucinations and other issues from the initial release, including instructions referencing internet data. The size ranges from 10K to 100K examples, and the cleaning repository is linked on GitHub.

What you can do with it

  • SFT for small language models
  • LoRA instruction training
  • Instruction-following evaluation
Tasks
Text generation
Languages
English
Size
10K–100K rows
Downloads this month
26.1K
Likes
907

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("yahma/alpaca-cleaned")