Salesforce Time Series Corpus
4.5 million time series and 230 billion data points for pre-training forecasting foundation models.
- Time series forecasting
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
4.5 million time series and 230 billion data points for pre-training forecasting foundation models.
Over 10,000 hours of real-world recordings of craftsmanship, cooking, and construction work with IMU sensor data.
A large corpus of tokenized time series (10 to 100 million records) without a detailed card description.
A massive archive of minute OHLCV candles for 5,795 Chinese Class A stocks for financial time-series forecasting.
Large multimodal dataset of 1M to 10M examples for training robots in safe human interaction.
Synchronized video, audio, and 3D data from real home environments for embodied AI training.
A mix of expert RL demonstrations, image-caption pairs, and text for training a multimodal agent.
Billions of cleaned web documents from Internet Archive and Common Crawl in 191 world languages.
Over 720 hours of dual-arm robot manipulation recordings with text annotations for training robotic models.
Tabular, textual, and time-series scientific data from Tahoe-100M for analysis and agent queries.
First dataset with images, text, and time series for semantic communication and 6G tasks.
Collection of real-world time series from AutoGluon for testing forecasting models.
Teleoperation recordings of a robot with two five-fingered hands for training robotics models.
Over 5,000 hours of driving across 30 German cities with 6-camera video, telemetry, and environment labels.
Large collection of synchronized video and tactile sensor data for training robot manipulation.
Large robot-free bimanual manipulation corpus: video, poses, and synchronized sensors.
A large array of English video, tabular, and time-series data for training robot control systems.
6.18 million RGB+Depth frames and over a million text instructions for training dual-arm robotic manipulators.
192 hours of video and telemetry of bimanual robot manipulation for training control systems.
Four LIBERO datasets with Panda robot episodes, merged and converted to the LeRobot format.
Over one million frames and 34 thousand video recordings of 24 tasks performed by a Franka robot in LeRobot format.
Multimodal logs of a physical dual-arm YAM robot with success and failure annotations.
Nearly 96 thousand episodes and 27.6 million frames with video, tables, and time series in parquet and LeRobot formats.
Time series collections for training and evaluating Chronos forecasting models.