softadaptx 0.0.1

SoftAdapt: Techniques for Adaptive Loss Weighting of Neural Networks with Multi-Part Loss Functions

SoftAdaptX

This repository contains an updated implementation of the SoftAdapt algorithm( techniques for adaptive loss balancing of multi-tasking neural networks). This work continues on the awesome work of Ali Heydari ( see https://github.com/dr-aheydari/SoftAdapt) and aims to have a more streamlined and versatile implementation of SoftAdapt.

Installation

Using pip

SoftAdaptX is officially released on PyPI. To install SoftAdaptX with pip:

pip install softadaptx

Local installation

SoftAdapt now uses Poetry for dependency management. To install SoftAdapt with Poetry:

1. First, make sure you have Poetry installed:

curl -sSL https://install.python-poetry.org | python3 -

2. Then, you can install SoftAdapt directly from GitHub:

poetry add git+https://github.com/dr-aheydari/SoftAdapt.git

3. Or clone the repository and install locally:

git clone https://github.com/dr-aheydari/SoftAdapt.git
cd SoftAdapt
poetry install

Contributing

Contributions are welcome. Please follow these steps to contribute:

1. Fork the repository on GitHub.
2. Create a new branch for your feature or bug fix.
3. Make your changes and commit them.
4. Push your changes to your fork.
5. Submit a pull request to the main branch of the original repository.