GraphMuse v0.0.5

GraphMuse Release 0.0.5

This minor release only builds the wheels for more recent Python version.
No code changes were applied.

GraphMuse Release 0.0.4

New Release 0.0.4

This latest release contains some updates and bug fixes

Updates

• Pitch Spelling model update to contain Spelling and Key encoders
• Support for PyG in Memory dataset
• New classes for Hybrid models
• Load midi from URL and return score graph
• Updated utilities for training models for voice separation
• Included minimal feature extraction from scores
• Improved memory handling during graph creation

Bug Fixes and Testing

• Bug fix of Hierarchical GraphSAGE last layer
• Added tests for Readme scripts
• Length of graphs for sampling only depends on notes

Happy graph coding!

GraphMuse Release 0.0.3

GraphMuse Release 0.0.3

This new version addresses minor errors, fixed examples, and easier installation.

Bump version for minor Fixes

Release 0.0.2

Minor fixes for better packaging with pip.

GraphMuse First Release 0.0.1

GraphMuse Release Notes

Version 1.0.0 - Initial Release

Release Date: July 18, 2024

Highlights

• Core Functionality:

  • Introduction of core graph-based music representations using PyTorch Geometric Data and HeteroData classes.
  • Sampling strategies for Music Score Graphs.
  • Dataloaders and Graph Creation classes.
  • Integration with the Partitura library for parsing symbolic music.

• Graph Neural Networks:

  • Implemented MusGConv, NoteGNN, MeasureGNN, BeatGNN, MetricalGNN, and HybridGNN models.

• Utilities:

  • Dataloader for sampling and batching music graphs.
  • Accelerated graph creation for music scores.

Installation

• Available via pip.
• Detailed installation steps for dependencies provided.

Documentation

• Examples and usage instructions available in the README.

Future Improvements

• Planned enhancements include additional graph convolutional layers and extended dataset support.

For more information, visit the GraphMuse GitHub repository.