flow-forecast
Deep learning PyTorch library for time series forecasting and anomaly detection.
Deep learning framework featuring transformers, attention models, and LSTMs for time series data.
Flow Forecast is an open-source PyTorch framework designed for time series forecasting, classification, and anomaly detection. It is built for data scientists and quantitative analysts who need state-of-the-art neural network architectures for predictive modeling.
What it is
The library provides implementations of modern time series models including various transformer architectures, Informer, DeepAR, DLinear, and traditional LSTMs and GRUs. It features cloud provider integration for Google Cloud Platform, experiment tracking with Weights and Biases, and model serving capabilities.
Who it's for
It suits machine learning engineers and financial data scientists looking to implement deep learning models for sequence data and risk forecasting. It is not suitable for analysts seeking simple statistical models like ARIMA without deep learning dependencies.
Key features
State-of-the-art Architectures
Includes implementations of Transformers, Informer, DeepAR, DA-RNN, and TSMixer for time series analysis.
Cloud and Tracking Integration
Comes with built-in integrations for Google Cloud Platform and Weights and Biases experiment tracking.
Multiple Task Support
Supports forecasting, classification, and anomaly detection tasks out of the box.
Up and running in 3 steps
Our honest verdict
What's great
- Comprehensive collection of transformer-based time series models.
- Native integration with Google Cloud Platform and Weights and Biases.
- Supports both classification and anomaly detection alongside forecasting.
- Actively maintained by the CoronaWhy task-ts team with ongoing model updates.
Where it falls short
- Requires significant Python and deep learning expertise to configure and tune.
- Heavy dependency on PyTorch and external cloud infrastructure for scaling.
- Documentation is spread across wikis, Jira spaces, and external repositories.
Pricing & license
Free under the GPL-3.0 licence. The only cost is running your own compute infrastructure and cloud services.
Open remote AI-training roles
874 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
What models are supported in Flow Forecast?
It supports Vanilla LSTM, Full Transformer, Informer, DeepAR, DLinear, Crossformer, TSMixer, and several attention-based models.
Can I integrate this with cloud platforms?
Yes, the library features specific integrations for Google Cloud Platform and Weights and Biases.
