Skip to content

Dataset · inria-soda

OpenML tabular dataset benchmark collection

inria-soda/tabular-benchmark

A collection of real tabular datasets from OpenML for comparing classification and regression algorithms.

What's inside

The dataset is a curated collection of tabular tasks from the OpenML repository with a total volume of 10 million to 100 million rows. It covers regression and classification tasks on numerical and categorical features. The compilation was created by researchers from Inria for standardized benchmarking of machine learning models.

What you can do with it

  • Comparing neural networks for tabular data with gradient boosting
  • Validating tabular classification and regression pipelines
Tasks
Tabular classification, Tabular regression
Size
10M–100M rows
Downloads this month
106.2K
Likes
52

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("inria-soda/tabular-benchmark")