v2.0.0

Breaking Changes

• Transform System: Complete refactor to PyTorch named tensors for GPU acceleration. Transforms now use ('channel', 'time') or ('batch', 'channel', 'time') tensor conventions
• Dataset Architecture: Refactored datasets module with layered architecture (WindowedDataset, DataModule, DatasetCreator)
• Models Restructure: Flattened model hierarchy from 4 levels to 2 (models/raul_net/v16.py). Removed deprecated models (V1, V4, V9 variants, classification), removed online/offline distinction
• Datatypes Package: Split datatypes.py into organized package structure
• Storage Format: Enforced .zip format for datasets, removed GPU Direct Storage support

Added

• SlidingWindowTransform base class for consistent transform implementations
• MLflow experiment tracking and visualization (myoverse/tracking.py)
• RAM caching and multiprocessing support in ContinuousDataset
• Lazy dataloader initialization for faster startup
• Cache pre-loading in main process before spawning workers
• Nested dataset storage structure (split/modality/task)
• ZipStore for faster dataset I/O on Windows
• py.typed marker for PEP 561 compliance

Changed

• Lazy import of heavy dependencies in datasets module for faster imports
• Modernized codebase to Python 3.12+ type hints and modern super() calls
• Ran ruff format across entire codebase
• Removed graph visualization from _Data class
• Removed Rich auto-coloring of numbers and text for cleaner output
• Moved emg_xarray and emg_tensor to datatypes module

Removed

• Unused utils module
• tests/workflow.py
• icecream dependency
• Deprecated model variants (V1, V4, V9, classification models)
• GPU Direct Storage (GDS) support

Fixed

• MedianFilter padding
• Tests updated to match simplified _Data API

Full Changelog: v1.1.6...v2.0.0

v1.1.6

Added

• GitHub Actions workflow for automated PyPI publishing using trusted publishing

Changed

• Fixed numcodecs dependency specification to >=0.13.0,<0.16 to avoid cbuffer_sizes import errors
• Renamed __add_data_to_dataset to _add_data_to_dataset in EMGDataset to fix Sphinx documentation generation

Fixed

• Resolved Sphinx autosummary import error caused by Python name mangling of double-underscore methods

Full Changelog: v1.1.5...v1.1.6

v1.1.5

Changed

• Fixed numba dependency specification from <=0.61.0 to >=0.61.0 for improved compatibility
• Updated documentation structure with enhanced README in docs/source
• Updated dependency lock file (uv.lock) to reflect current dependencies

Added

• .gitignore entry for version_* files to prevent tracking temporary version files

Full Changelog: v1.1.4...v1.1.5

v1.1.4

Added

• Added description to pyproject.toml

Changed

• Version bump from 1.1.3 to 1.1.4

Full Changelog: v1.1.3...v1.1.4

v1.1.3

Added

• Package __version__ attribute in myoverse/__init__.py that dynamically reads from pyproject.toml
• Version number included in logging timestamp prefix

Changed

• PyPI compatible dependency specifications for PyTorch with separate configurations for Linux and Windows
• Updated installation instructions in README for clarity on platform-specific PyTorch installations
• Enhanced GitHub links in project.urls section

Full Changelog: v1.1.2...v1.1.3

v1.1.2

Changed

• Updated README logo URL to use raw GitHub link for better cross-platform compatibility
• Revised Lightning to version 2.5.0.post0
• Removed PyQt5 and related packages from development dependencies
• Streamlined dependency list while maintaining compatibility

Full Changelog: v1.1.1...v1.1.2

v1.1.1

Changed

• Removed unnecessary string handling in EMGDataset class
• Mean and standard deviation are no longer nn.Parameter objects in RaulNetV17 model. They should have been regular tensors
• Fix version of lightning to 2.5.0.post0 in pyproject.toml to avoid bug with mlflow integration

Full Changelog: v1.1.0...v1.1.1

v1.1.0

Added

• RaulNetV17 model for decoding kinematics from EMG data
• Workflow module for EMG data processing, model training, and result visualization
• New reStructuredText files for data types and functions documentation
• Unit tests for EMG data augmentation filters and default datasets

Changed

• Refactored code for improved readability and consistency in optimizer configuration, visualization functions, and data class initializations
• Updated test data class to include dimensions when unchunked
• Downgraded zarr version from 3 to 2 in loader and supervised modules
• Refactored optimizer configuration and updated ground truth handling in training steps
• Refactored EMG and kinematics data classes to include dimensions when unchunked and streamlined chunk checking logic
• Enhanced kinematics visualization with interactive features and improved styling
• Updated zarr version and added mlflow and pyqt5 dependencies
• Renamed AveragingSpatialFilter to ApplyFunctionSpatialFilter in documentation
• Refactored spatial filters simplifying parameters and type hints
• Refactored SOSFrequencyFilter for easier usability
• Enhanced grid layout documentation and validation in EMGData class

Removed

• ElectrodeSelector and GridReshaper filters mentioned in documentation
• Unnecessary spatial filters

Full Changelog: v1.0.0...v1.1.0

v1.0.0

Added

• Examples for documentation
• Sections for Filters and Tutorials in README
• Example scripts for training models and dataset creation
• Enhanced documentation structure with autosummary for datasets and filters
• Unit tests for core functionality

Changed

• Bump version to 1.0.0 and add pydata-sphinx-theme dependency
• Updated model definitions to indicate archival status
• Replaced pytorch_lightning with lightning
• Enhanced documentation for VirtualHandKinematics class by adding reference to MyoGestic
• Refactored activation function documentation and improved loss calculation formatting
• Enhanced dataset creation and loading by improving EMG dataset handling
• Added empty data checks and refined augmentation pipelines
• Refactored dataset creation with new CustomDataClass
• Enhanced EMG dataset loading to support Zarr files with improved validation
• Enhanced filter configurations by adding input_is_chunked parameter
• Improved EMGAugmentation class with additional parameters and better documentation
• Improved spatial filter API
• Enhanced filter methods to accept additional keyword arguments
• Enhanced layout algorithm for EMG visualizations with topological sorting and improved node positioning

Full Changelog: v0.1.0...v1.0.0