https://github.com/daniel05155/reinforcement-learning-in-music-generation

An implementation of automated music composition using Deep Reinforcement Learning with PyTorch.
https://github.com/daniel05155/reinforcement-learning-in-music-generation

inverse-reinforcement-learning music-generation python pytorch reinforcement-learning

Last synced: 3 months ago
JSON representation

An implementation of automated music composition using Deep Reinforcement Learning with PyTorch.

• Host: GitHub
• URL: https://github.com/daniel05155/reinforcement-learning-in-music-generation
• Owner: daniel05155
• Created: 2024-06-07T12:17:38.000Z (12 months ago)
• Default Branch: master
• Last Pushed: 2024-09-09T08:17:00.000Z (9 months ago)
• Last Synced: 2025-01-16T13:49:27.685Z (5 months ago)
• Topics: inverse-reinforcement-learning, music-generation, python, pytorch, reinforcement-learning
• Language: Python
• Homepage:
• Size: 180 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Automated Music Composition Using Deep Reinforcement Learning

This repository is an implementation of automated music composition with Reinforcement Learning(RL) in PyTorch.

We design a novel deep reinforcement learning-based framework with the self-attention-based neural networks for long-sequence music generation, which differs from the conventional Generative Adversarial Network (GAN).

Additionally, we utilize  Deep Q Network (DQN)-based and Proximal Policy Optimization (PPO)-based techniquein Reinforcement Learning (RL) with the transformer model to generate pop-style piano music

sequentially.

## Model architecture

###  DQN-Baed Architecture

![Aritecture](AIRL.png)

###  PPO-Baed Architecture

![Aritecture](PPO_Architecture.png)

# Getting started

## Install Dependencies

```

pip install -r requirements.txt

```

## Dataset

We use the "AIlabs-Pop1K7" as training dataset, which can be download [here](https://github.com/YatingMusic/compound-word-transformer/tree/main).  

## Pre-trained Weights

Linear Transformer is our pre-trained weights of agent model, which can be downloaded [here](https://drive.google.com/file/d/1qRALPNx--GCzLJqrkEfP0bcm91jpkvcI/view?usp=sharing).

## Run the Codes

Our expertiment is based on the unconditional generation with Reinforcement Learning using PyTorch framework in Python.

* Pretrain: 

```

python agent_pretrain.py

```

* Training

```

python IRL_dqn_train.py

```

* Evaluation

```

## Generate midis songs in './gen_midis ##

python testing-no-type-cp.py

```

## Demo

Our generated midi files can be downloaded [here](https://drive.google.com/drive/folders/13HKoJLNTd_gKfzMb2OlZP4nKBRlLHfQ3?usp=sharing).