timeserio is the missing link between pandas, scikit-learn and keras. It simplifies building end-to-end deep learning models - from a DataFrame through feature pipelines to multi-stage models with shared layers. While initially developed for tackling time series problems, it has since been used as a versatile tool for rapid ML model development and deployment.
Loosing track of big networks with multiple inputs and outputs? Forgetting to freeze the right layers?
Struggling to re-generate the input features? timeserio can help!
Please see the official documentation on how to get started.
- Enable encapsulated, maintainable and reusable deep learning models
- Feed data from
pandasthroughscikit-learnfeature pipelines to multiple neural network inputs - Manage complex architectures, layer sharing, partial freezing and re-training
- Provide collection of extensible building blocks with emphasis on time series problems
pip install timeserio, or install from source - pip install -e .
See Getting Started
We welcome contributions and enhancements to any part of the code base, documentation, or tool chain.
See CONTRIBUTING.md for details on setting up the development environment, running tests, etc.