Skip to content

Dataset · bigscience

P3: prompted NLP tasks

bigscience/P3

Collection of English NLP datasets with prompt templates for inputs and targets.

What's inside

P3 (Public Pool of Prompts) is a collection of English datasets covering various NLP tasks. Examples are provided with input and target templates that convert fields into natural language sequences. Size is 100M-1B, modality is text.

What you can do with it

  • Train an instruction-following model
  • Evaluate zero-shot and few-shot performance
  • Investigate the impact of prompt templates
Tasks
Other
Languages
English
Size
100M–1B rows
Downloads this month
76.1K
Likes
245

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("bigscience/P3")