Skip to content

Dataset · Salesforce

Salesforce Time Series Corpus

Salesforce/GiftEvalPretrain

4.5 million time series and 230 billion data points for pre-training forecasting foundation models.

What's inside

The collection combines 71 univariate and 17 multivariate time series datasets from seven distinct domains with 13 sampling frequencies. In total, it contains 4.5 million series and 230 billion data points. The dataset is filtered to avoid overlap with the GIFT-Eval benchmark test set to prevent data leakage.

What you can do with it

  • Pre-training foundation models for time series forecasting
  • Fair model evaluation on the GIFT-Eval leaderboard without data leakage
  • Testing forecasting algorithms on series with varying frequencies
Tasks
Time series forecasting
Size
1M–10M rows
Downloads this month
491.1K
Likes
56

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Salesforce/GiftEvalPretrain")