leaf-1/pyproject.toml at main · csglab/leaf-1 · GitHub

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
[build-system]
requires = ["setuptools>=68", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "leaf1"
version = "0.1.0"
description = "LEAF-1: semantic fragment representations for coordinate-free analysis of genomics data"
readme = "README.md"
requires-python = ">=3.10"
license = { text = "PolyForm Noncommercial License 1.0.0" }
authors = [
  { name = "Hamed Heydari", email = "hheydarii@gmail.com" },
]
keywords = ["cfDNA", "scATAC", "transformer", "fragmentomics", "MLM", "DNA language model"]
classifiers = [
  "Programming Language :: Python :: 3",
  "Programming Language :: Python :: 3.10",
  "Programming Language :: Python :: 3.11",
  "Operating System :: POSIX :: Linux",
  "Topic :: Scientific/Engineering :: Bio-Informatics",
  "Intended Audience :: Science/Research",
]

# Versions mirror the paper-era runtime. Keep upper bounds tight for reproducible
# extraction: newer torch/transformers combinations may select CUDA runtimes that
# exceed local drivers or route DistilBERT through different attention kernels.
dependencies = [
  "torch>=2.1,<2.2",
  "transformers>=4.36,<4.38",
  "tokenizers>=0.15,<0.16",
  "accelerate>=0.25,<0.27",
  "evaluate>=0.4",
  "scikit-learn>=1.3",
  "numpy>=1.24,<2",
  "pandas>=1.5",
  "h5py>=3.9",
  "tqdm>=4.65",
  # Downstream MIL + mean-pool deps
  "pyyaml>=6",
  "threadpoolctl>=3",
  "joblib>=1.3",
  "matplotlib>=3.7",
  "seaborn>=0.12",
]

[project.optional-dependencies]
test = [
  "pytest>=8",
  "coverage>=7",
]

[project.scripts]
leaf1-pretrain = "leaf1.train.pretrain:main"
leaf1-extract = "leaf1.infer.extract_embeddings:main"
leaf1-mean-pool-aggregate = "leaf1.downstream.mean_pool.aggregate:main"
leaf1-mean-pool-train = "leaf1.downstream.mean_pool.train:main"
leaf1-mil = "leaf1.downstream.mil.train:main"
leaf1-mil-predict = "leaf1.downstream.mil.inference:main"
leaf1-bed-preprocess = "leaf1.util.bed_preprocess:main"
leaf1-reshape-layer = "leaf1.util.reshape_layer:main"
leaf1-validate-extract = "leaf1.util.validate_extract:main"

[project.urls]
Homepage = "https://github.com/csglab/leaf-1"
Issues = "https://github.com/csglab/leaf-1/issues"

[tool.setuptools]
include-package-data = true

[tool.setuptools.packages.find]
where = ["."]
include = ["leaf1*"]
exclude = ["tests*", "scripts*", "assets*"]

[tool.setuptools.package-data]
# Ship the tokenizer with the wheel so users have a reproducible default.
# Large assets live in the source checkout through Git LFS, not inside wheels.
# Also ship the MIL sweep-grid YAMLs.
leaf1 = [
  "data/tokenizer.json",
  "downstream/mil/configs/*.yaml",
]

[tool.pytest.ini_options]
testpaths = ["tests"]
addopts = "-ra"
filterwarnings = [
  "ignore::DeprecationWarning",
  "ignore::FutureWarning",
]