EigenLab is an open-source research laboratory that explores the integration of advanced mathematical concepts into deep neural network architectures. It provides a flexible framework for developing and experimenting with innovative models that leverage principles from fields such as algebra, differential geometry, and physics.
EigenLab aims to bridge the gap between advanced mathematics and deep learning, enabling researchers and developers to explore the potential of incorporating new or improving on the existing mathematical structures related to the design and training of neural networks. The project focuses on creating an environment that encourages experimentation, fosters collaboration, and facilitates the development of cutting-edge models.
- Covariant Derivative Transformer: An example implementation of a transformer-based language model that incorporates the concept of covariant derivatives.
- Modular Architecture: EigenLab provides a flexible and extensible architecture for building deep learning models that leverage advanced mathematical principles.
- Model Repository: A dedicated directory (
models/) for storing trained models, allowing easy access and version control.
-
Clone the EigenLab repository:
git clone https://github.com/zwimpee/eigenlab.git cd eigenlab -
Install the required packages:
pip install -r requirements.txt
-
Install the package:
pip install e .
After installing EigenLab, you can use it to experiment with advanced mathematical principles in neural networks. Here is a basic example of how to use it:
from eigenlab import CovariantDerivativeTransformer
# Define the model
model = CovariantDerivativeTransformer(d_model=512, nhead=8, num_layers=6)
# Load some data
data = ...
# Train the model
model.train(data)EigenLab is an open-source project and we welcome contributions of all sorts. There are many ways to help, from reporting issues, contributing code, and helping us improve our community. Check out our contributing guide for more details.
This project is licensed under the terms of the MIT license. See the LICENSE file for more info.