Skip to content

Dataset · nvidia

OpenMathInstruct-2: 14M Math Pairs

nvidia/OpenMathInstruct-2

14 million math problem-solution pairs generated by Llama 3.1 405B for instructional fine-tuning.

What's inside

A dataset for math instruction tuning: 14 million pairs of problems and solutions. Problems from the GSM8K and MATH training sets were used to generate chain-of-thought solutions and new problem-solution pairs. Everything was generated by the Llama3.1-405B-Instruct model; the language is English.

What you can do with it

  • SFT of a mathematical LLM
  • Fine-tuning a model to solve GSM8K and MATH
  • Evaluating chain-of-thought generation
Tasks
Question answering, Text generation
Languages
English
Size
10M–100M rows
Downloads this month
116K
Likes
271

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/OpenMathInstruct-2")