v1.3.0

We are excited to announce the release of v1.3.0! This update includes several enhancements and usability improvements.

Highlights

UTMOSv2 now supports audio that is already in memory! This would be useful for many use cases, including employing UTMOSv2 as a reward model or integrating UTMOSv2 into custom pipelines.
See the corresponding issue and PR for more details:

• #81
• #82

What's Changed

• Add py.typed marker file for PEP 561 compliance by @kAIto47802 in #68
• Relax torch version requirement by @kAIto47802 in #72
• remove audio file whitelist, fallback to numpy load if audio can't be read by librosa backend by @tkanarsky in #75
• Update ci conditions by @kAIto47802 in #76
• Support Python 3.13 by @kAIto47802 in #77
• Update type annotations by @kAIto47802 in #80
• Support audio that is already in memory by @kAIto47802 in #82
• Refactor core modules by @kAIto47802 in #83
• Update README by @kAIto47802 in #84
• Add test_predict by @kAIto47802 in #85
• Add sr argument for the user input by @kAIto47802 in #86
• Update README.md by @kAIto47802 in #87
• Update README.md by @kAIto47802 in #88
• Fix device mismatch in torchaudio Resample when input is on CUDA by @kAIto47802 in #90
• Introduce uv by @kAIto47802 in #91
• Bump up to v1.3.0 by @kAIto47802 in #92

New Contributors

• @tkanarsky made their first contribution in #75

Full Changelog: v1.2.1...v1.3.0

v1.2.1

This is the release note of v1.2.1.

What's Changed

• Skip loading indivisual component weight when the fusion model weight is specified by @kAIto47802 in #64
• Bump up to v1.2.1 by @kAIto47802 in #65

Full Changelog: v1.2.0...v1.2.1

v1.2.0

This is the release note of v1.2.0.

What's Changed

• Add optional utility remove_silent_section for users by @kAIto47802 in #57
• Update optional utility remove_silent_section by @kAIto47802 in #58
• Bump up to v1.2.0 by @kAIto47802 in #61

Full Changelog: v1.1.1...v1.2.0

v1.1.1

This is the release note of v1.1.1. The pre-trained weights are moved to Hugging Face Models.

What's Changed

• Update import order by @kAIto47802 in #52
• Move pretrained models to HF Models by @kAIto47802 in #53

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

v1.1.0

We are excited to announce the release of v1.1.0! This update includes several enhancements and usability improvements.

The pre-trained weights previously hosted on GitHub have been removed, and the new code now uses those stored on 🤗Hugging Face. Please download and use them from there from now on:

Also we attached the poster presented at IEEE SLT 2024, which can be used for a quick overview:

What's Changed

• Update README.md by @kAIto47802 in #36
• Add CITATION.cff by @kAIto47802 in #37
• Add SLT Poster by @kAIto47802 in #38
• Fix name order in citation by @kAIto47802 in #39
• Remove ckpt by @Wataru-Nakata in #41
• Update VSCode settings by @kAIto47802 in #43
• Add config type by @kAIto47802 in #42
• Update README and remove .gitattributes by @kAIto47802 in #44
• Set disallow_untyped_defs=true and add missing type annotations by @kAIto47802 in #45
• Update download utils by @kAIto47802 in #46
• Remove unused save utils by @kAIto47802 in #47
• Rename BaseDataset to _BaseDataset for internal use by @kAIto47802 in #48
• Add tests by @kAIto47802 in #49
• Update torch installation for the CPU-only version in CI by @kAIto47802 in #50
• Bump up to v1.1.0 by @kAIto47802 in #51

New Contributors

• @Wataru-Nakata made their first contribution in #41

Full Changelog: v1.0.1...v1.1.0

v1.0.1

This is the release note of v1.0.1.

What's Changed

• Add __future__.annotations to newly updated moduels by @kAIto47802 in #32
• Add __future__.annotations to utmosv2/dataset/_schema.py by @kAIto47802 in #34
• Bump up to v1.0.1 by @kAIto47802 in #35

Full Changelog: v1.0.0...v1.0.1

v1.0.0

We are excited to announce the release of v1.0.0! This update includes several enhancements, bug fixes, and new features aimed at making the project more user-friendly and efficient.

What's Changed

• Convert file_path from str to Path by @kAIto47802 in #4
• Fix error when input directory contains a single wav file by @kAIto47802 in #5
• Fix window length by @kAIto47802 in #10
• Update issue and PR templates by @kAIto47802 in #11
• Refactor modules by @kAIto47802 in #12
• Introduce mypy by @kAIto47802 in #13
• Update workflows by @kAIto47802 in #14
• Use quotes for type annotations by @kAIto47802 in #15
• Reduce dependencies for Inference by introducing lazy import by @kAIto47802 in #17
• Improve documentation for wandb API key setup and PEP 660 support by @kAIto47802 in #18
• Configure package discovery in setuptools to ensure correct installation via pip by @kAIto47802 in #19
• Dynamically set version from __version__ in utmosv2 package by @kAIto47802 in #20
• Refactor dataset handling with BaseDataset class and add transform type annotations by @kAIto47802 in #21
• Refactor settings to make it non-public by @kAIto47802 in #22
• Add user-friendly model creation and prediction functionality by @kAIto47802 in #23
• Fix naming convention inconsistency in optional dependencies in pyproject.toml by @kAIto47802 in #24
• Update docs for user-friendly model creation and prediction functionality by @kAIto47802 in #25
• Remove pandas dependencies by lazy import by @kAIto47802 in #26
• Add docstrings by @kAIto47802 in #27
• Update instructions for optional dependencies by @kAIto47802 in #28
• Update paper information by @kAIto47802 in #29
• Update quickstart.ipynb instructions to reflect new user-friendly API by @kAIto47802 in #30
• Bump up version to v1.0.0 by @kAIto47802 in #31

New Contributors

• @kAIto47802 made their first contribution in #4

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