tsai
tsai is an open-source state-of-the-art deep learning library for time series and sequence analysis built on PyTorch and fastai. It provides advanced tools for classification, regression, forecasting, and imputation tasks. The package includes a comprehensive collection of models ranging from classic architectures like LSTM, GRU, and MLP to modern transformer-based solutions such as PatchTST and TabFusionTransformer. It features a vast repository of over 200 datasets for unified, multivariate, and forecasting applications. Key capabilities include sklearn-style pipeline transforms, walk-forward cross-validation, and optimized memory usage for accurate long-term forecasting. The library requires Python 3.10 or newer and is distributed via PyPI. Designed for data scientists and researchers, tsai simplifies the implementation of complex deep learning workflows while supporting PyTorch 2.0.