Metadata-Version: 2.2
Name: audiozen
Version: 2025.1.13.1
Summary: Audio ZEN is a library for audio/speech signal processing.
Requires-Python: >=3.12
Description-Content-Type: text/markdown
Requires-Dist: einops
Requires-Dist: librosa
Requires-Dist: simple_parsing
Requires-Dist: torchinfo
Requires-Dist: tensorboard
Requires-Dist: huggingface-hub[hf-transfer]
Requires-Dist: pyyaml
Requires-Dist: numpy
Requires-Dist: numba>=0.60
Requires-Dist: ipykernel
Requires-Dist: tqdm
Requires-Dist: pandas
Requires-Dist: tabulate
Requires-Dist: matplotlib
Requires-Dist: joblib
Requires-Dist: psutil
Requires-Dist: torch>=2.5.1
Requires-Dist: torchvision>=0.20.1
Requires-Dist: torchaudio>=2.5.1
Provides-Extra: demo-crf-wp2
Requires-Dist: plotly>=5.24.1; extra == "demo-crf-wp2"
Requires-Dist: streamlit>=1.41.1; extra == "demo-crf-wp2"
Provides-Extra: dev
Requires-Dist: bump-my-version; extra == "dev"

---
license: mit
language:
- en
pipeline_tag: audio-to-audio
---
# AudioZEN

## Prerequisites

```bash
# Install uv for speed up virtual environment creation and management
uv venv -p 3.12 venv/torch251_cu124_py312
source venv/torch251_cu124_py312/bin/activate

# Install the package
uv pip install -e .

# cd to the model directory
uv pip install -r /path/to/requirements.txt
```

## Features

- [x] Gradient accumulation
- [x] Multi-node training
- [x] BF16 support
- [x] Learning rate warmup
- [x] Learning rate decay
  - [x] Linear decay

## Prerequisites

```shell
rsync -avPxH --no-g --chmod=Dg+ /home/xhao/proj/audiozen xhao@10.21.4.91:/home/xhao/proj/audiozen --exclude="*.git" --exclude="*.egg-info" --exclude="*.egg" --exclude="*.pyc" --exclude="*.log" --exclude="*.npy"
```

- How to split the repo into read-only standalone repos? Check out [Monorepo-Management](https://github.com/haoxiangsnr/audiozen/wiki/Monorepo-Management)

## Git LFS

If files are too large and/or change frequently, consider using [Git LFS](https://git-lfs.github.com/).

```shell
git lfs install
git lfs track "*..."
git add .gitattributes
```

For ipynb files, we donnot need to track them as they are not large files and they are not changed frequently.
