Catasta is a Python library designed to simplify the process of Machine Learning model experimentation. It encapsulates the complexities of model training, evaluation, and inference in a very simple API.
Warning
๐ง Catasta is in early development ๐ง
Expect breaking changes on every release until v1.0.0
is reached.
Also, The documentation and examples for the library are under development.
Catasta is a very simple and easy to use package.
Catasta offers a variety of pre-implemente Machine Learning models. All models are single-scripted, so feel free to copy and paste them anywhere.
For regression:
- Approximate Gaussian Process
- Transformer
- Transformer with FFT
- Mamba
- Mamba with FFT
- FeedForward Neural Network
For classification:
- Convolutional Neural Network
- Transformer
- Transformer with FFT
- Mamba
- Mamba with FFT
- FeedForward Neural Network
Provides an easy way to import the data contained in directories.
Let's you apply transformations to the data when its loaded to a dataset, such as window sliding, normalization...
Scaffolds are where models and datasets are integrated for training, handling both training and evaluation.
Catasta supports and plans to support the following Machine Learning tasks:
- SISO Regression
- MISO Regression
- Image Classification
- Signal Classification
- Binary Classification
- Probabilistic Regression and Classification
Takes a trained model and handles the inference task.
Catasta is available as a PyPi package:
pip install catasta
Clone the repository
git clone https://github.com/vistormu/catasta
and install the dependencies
pip install -r requirements.txt