#136136th of 168 in ML frameworks and libraries
tsai
PyTorch and fastai library for time series classification and forecasting
- Stars
- 6.1k
- License
- Apache-2.0
- Last commit
- Jun 2026
- Last release
- May 2026
- Language
- Jupyter Notebook
6,004 on 20 Apr 20266,144 on 9 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
tsai is a Python package built on PyTorch and fastai for training deep learning models on time series: classification, regression, forecasting and imputation. It bundles about 30 architectures (InceptionTime, TST, PatchTST, MiniRocket, TCN and others), dataset downloaders and tutorial notebooks. Models take a 3D array of samples x variables x sequence length and export to a pickled learner for inference.
Who it is for: ML engineers and researchers modeling time series in Python
Strengths
- Ships many time series architectures, including PatchTST, InceptionTime, TST and MiniRocket
- Downloads 128 univariate and 30 multivariate classification datasets, plus regression and forecasting sets
- Plus models auto-configure the output head to match the forecasting target shape
- Tutorial notebooks, including a Colab intro to time series classification
Weaknesses
- Tied to the fastai training loop and API, which adds a learning curve
- ROCKET models need sktime installed separately or via the extras install
- Requires Python 3.10 or newer; conda channel is no longer updated
- Library only: no serving layer, UI or deployment tooling
What it needs
- GPU optional
- Compose
- Needs PyTorch, fastai, Python 3.10+, sktime (optional), tsfresh (optional)
- Models: PatchTST, InceptionTime, TST, TSiT, MiniRocket
What changed
- v1.0.0 ↗ (opens in a new tab)Breaking
“PyPI users: bump your Python to 3.10+ before pip install -U tsai. On Python 3.9 the upgrade will fail to resolve”
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Quick answers
Is tsai open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Can you run tsai with Docker?
Yes. Its repository has a Docker setup: a Dockerfile, a Compose file or a published image.
Does tsai need a GPU?
No. It runs on a CPU and can use a GPU when you have one.
Also in ML frameworks and libraries
See all 168| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd | 93 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 93 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and training scripts | 93 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 92 out of 100 |