pyproject.toml

[project]
name = "mini-codec"
version = "0.0.1"
description = "A from-scratch implementation of a neural audio codec (Encodec/SoundStream-style) with a focused improvement experiment."
readme = "README.md"
requires-python = ">=3.10"
authors = [{ name = "Ruiting Ma" }]
license = { text = "MIT" }

dependencies = [
    "torch>=2.2",
    "torchaudio>=2.2",
    "numpy>=1.26",
    "scipy>=1.11",
    "einops>=0.7",
    "soundfile>=0.12",
    "librosa>=0.10",
    "pyyaml>=6.0",
    "tqdm>=4.66",
    "wandb>=0.16",
    "tensorboard>=2.16",
    "matplotlib>=3.8",
    "pyarrow>=14",
]

[project.optional-dependencies]
dev = [
    "pytest>=7.4",
    "pytest-cov>=4.1",
    "ruff>=0.3",
    "ipykernel>=6.29",
    "jupyter>=1.0",
]
metrics = [
    # Objective audio quality metrics. Optional because they pull in heavier deps.
    "pesq>=0.0.4",
    "pystoi>=0.4",
]
hf = [
    # HuggingFace fallback for the perceptual loss backend; only required
    # when configs/exp_b_*.yaml's loss.perceptual.backend is 'huggingface'
    # (e.g. on AutoDL CN regions where the torchaudio Meta CDN is gated).
    "transformers>=4.30",
]

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[tool.hatch.build.targets.wheel]
packages = ["src"]

[tool.ruff]
line-length = 100
target-version = "py310"

[tool.ruff.lint]
select = ["E", "F", "I", "B", "UP"]

[tool.pytest.ini_options]
testpaths = ["tests"]
addopts = "-q"