language:
- en
license: mit
datasets:
- aligned_bach_chorales
tags:
- gan
- music-generation
- pytorch
model-index:
- name: Experimental GAN for Bach-like Textures
results:
- task:
name: Music Generation
type: music-generation
dataset:
name: Aligned Bach Chorales Dataset
type: aligned_bach_chorales
metrics:
- name: Quality of Generated Music
type: subjective
value: Experimental
Experimental GAN for Bach-like Textures
Description
This repository contains an experimental Generative Adversarial Network (GAN) model designed to generate Bach-like textures. The model is based on the Aligned Bach Chorales Dataset, which can be found here.
License
This project is licensed under the MIT License.
Model Overview
The GAN consists of two parts: a generator and a discriminator. Both models were trained on the Aligned Bach Chorales Dataset, which represents Bach chorales in a binary matrix format.
Installation and Requirements
To use this model, you need to have PyTorch version 2.1.0+cu121 installed. The dataset loader, model structure, and training scripts are all included in the torch_binary2_v1 notebook.
Usage
- Clone the Repository:
git clone https://huggingface.co/Egorp/j_s_gan
- Load the Models:
- Use torch_binary2_v1 notebook to load the provided state dictionaries for both the generator and the discriminator.
- Convert Binary Matrices to MIDI:
- Utilize the
binary_to_midinotebook included in this repository to convert the binary matrices generated by the GAN into MIDI format.
Dataset
The dataset used for training this model can be found in the folder named "np_convert". This folder contains binary representations of Bach chorales.
Contributing
Contributions to this project are welcome. Please feel free to submit issues and pull requests.
Contact
For any queries or discussions regarding this project, please open an issue in this repository.
Acknowledgements
Special thanks to the creators and maintainers of the Aligned Bach Chorales Dataset for providing the data used to train this model.