torchcast: A PyTorch Library for Time Series Analysis

torchcast is a simple module for time series analysis using deep learning. It can be downloaded from:

https://github.com/TheoremEngine/torchcast

Or installed by:

pip install torchcast

torchcast is still under heavy construction. The intent for torchcast, at least in the near term, is to focus on data fetching and preprocessing. torchcast currently supports automatic ingestion of time series classification datasets from the UCR/UEA Archive, time series forecasting datasets from the Monash Archive, and the LSTF benchmark. Most of these datasets are small enough to hold in memory, where we standardize on a 3-dimensional (Batch, Channel, Time) arrangement. We intend to add additional datasets in the future.

Contents:

• Data Submodule
• Datasets Submodule
• nn Submodule
• Utilities Submodule