Repository navigation
Expand file tree
/
Copy pathpyproject.toml
More file actions
112 lines (103 loc) · 2.7 KB
/
Copy pathpyproject.toml
File metadata and controls
112 lines (103 loc) · 2.7 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
[project]
name = "rubatocodec"
version = "1.0.0"
description = "RubatoCodec: Budgeted Adaptive Speech Tokenization with Streaming Decoding"
readme = "README.md"
# RubatoCodec is Apache-2.0; the distribution also includes MIT-licensed code.
license = "Apache-2.0 AND MIT"
license-files = ["LICENSE", "incl_licenses/LICENSE*"]
requires-python = ">=3.12"
# Core dependencies cover inference only (rubatocodec package).
# Training / evaluation tooling is behind the optional extras below.
dependencies = [
"einops>=0.8.0",
"huggingface-hub>=0.24.0",
"numpy>=2.0.0",
"omegaconf>=2.3.0",
"safetensors>=0.4.0",
"soundfile>=0.12.0",
# torch/torchao/torchtune are pinned as a verified working set (torchao 0.18
# is incompatible with torchtune 0.6: torchao.dtypes.nf4tensor moved).
"torch>=2.12.1,<2.13",
"torchao>=0.17.0,<0.18",
"torchaudio>=2.11.0",
"torchtune>=0.6.0,<0.7",
"tqdm>=4.66.0",
"transformers>=4.44.0",
"vector-quantize-pytorch>=1.17.0",
]
[project.optional-dependencies]
# Training stack (training/ directory): PyTorch Lightning + Hydra + data pipeline
train = [
"datasets>=3.0.0",
"hydra-core>=1.3.2",
"librosa>=0.10.0",
"matplotlib>=3.8.0",
"omegaconf>=2.3.0",
"pesq>=0.0.4",
"pystoi>=0.4.1",
"pytorch-lightning>=2.4.0",
"pyyaml>=6.0",
"torchcodec>=0.14.0",
"torchdata>=0.9.0",
"wandb>=0.28.0",
]
# Evaluation stack (eval/ directory): objective metrics
eval = [
"datasets>=3.0.0",
"funasr>=1.1.0",
"jiwer>=3.0.0",
"librosa>=0.10.0",
"matplotlib>=3.8.0",
"omegaconf>=2.3.0",
"onnxruntime-gpu>=1.28.0",
"pandas>=2.0.0",
"pesq>=0.0.4",
"pyarrow>=15.0.0",
"pystoi>=0.4.1",
"s3prl>=0.4.17",
"speechbrain>=1.0.0",
"torchcodec>=0.14.0",
"torchmetrics>=1.4.0",
"utmosv2 @ git+https://github.com/sarulab-speech/UTMOSv2.git",
"zhconv>=1.4.0",
"zhon>=2.0.0",
]
[build-system]
requires = ["setuptools>=77.0.3", "wheel"]
build-backend = "setuptools.build_meta"
# Only the inference API (rubatocodec/) is distributed as a package;
# training/eval/scripts are development directories and are not packaged.
[tool.setuptools.packages.find]
include = ["rubatocodec*"]
[tool.ruff]
line-length = 100
target-version = "py312"
[tool.ruff.lint]
select = [
"E",
"W",
"F",
"I",
"B",
"C4",
"UP",
]
ignore = [
"E501",
"B008",
"C901",
]
[tool.ruff.lint.per-file-ignores]
"__init__.py" = ["F401"]
[tool.ruff.lint.isort]
known-first-party = ["rubatocodec", "training", "eval", "examples"]
[tool.ruff.format]
quote-style = "double"
indent-style = "space"
skip-magic-trailing-comma = false
line-ending = "auto"
[dependency-groups]
dev = [
"ruff>=0.15.5",
]