affex/pyproject.toml at main · norsage/affex · 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
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
[project]
name = "affex"
version = "0.1.0"
description = "Physics-informed DL for protein-protein binding free energy estimation"
requires-python = ">=3.12"
dependencies = [
    "pandas>=2.3.2",
    "scipy>=1.16.1",
    "gemmi>=0.7.3",
    "aim>=3.29.1",
    "hydra-core>=1.3.2",
    "loguru>=0.7.3",
    "jaxtyping>=0.3.3",
    "lightning>=2.5.4",
]

[project.optional-dependencies]
# Two mutually-exclusive install profiles for torch + PyG (see docs/release-migration.md §2).
# macOS/CPU:  uv sync --extra cpu     Linux/CUDA: uv sync --extra cu128
cpu = [
    "torch>=2.8.0,<2.9",
    # <2.8: torch-geometric 2.8 routes kNN through pyg-lib (no macOS-arm wheel) and ignores
    # torch-cluster. 2.7.x keeps the torch-cluster backend EXP-043 was trained on. See migration §2.
    "torch-geometric>=2.6.1,<2.8",
    "torch-cluster>=1.6.3",
    "torch-scatter>=2.1.2",
    "torch-sparse>=0.6.18",
]
# cu128 is a Linux-only profile. The markers keep the macOS slice of uv's universal
# resolution from requesting these and falling back to PyPI sdists (which build-from-source
# and need torch at build time). On Linux they resolve to data.pyg.org prebuilt wheels.
cu128 = [
    "torch>=2.8.0,<2.9 ; sys_platform == 'linux'",
    "torch-geometric>=2.6.1,<2.8 ; sys_platform == 'linux'",
    "torch-cluster>=1.6.3 ; sys_platform == 'linux'",
    "torch-scatter>=2.1.2 ; sys_platform == 'linux'",
    "torch-sparse>=0.6.18 ; sys_platform == 'linux'",
]
jupyter = [
    "jupyter>=1.1.1",
    "matplotlib>=3.10.8",
    "notebook>=7.3.3",
    "plotly>=6.3.0",
]

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

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

[dependency-groups]
dev = [
    "pytest",
    "ruff",
    "pyright",
    "kaleido>=1.2.0",
]

[tool.uv]
# cpu and cu128 extras are mutually exclusive — uv resolves a separate lock branch for each.
conflicts = [[{ extra = "cpu" }, { extra = "cu128" }]]

# torch itself: pull from the matching PyTorch index per extra.
[[tool.uv.index]]
name = "pytorch-cpu"
url = "https://download.pytorch.org/whl/cpu"
explicit = true
[[tool.uv.index]]
name = "pytorch-cu128"
url = "https://download.pytorch.org/whl/cu128"
explicit = true

# PyG compiled extensions: data.pyg.org is a flat (find-links style) HTML index, one per torch+cuda.
# For a different CUDA target, swap cu128 -> cu126/cu129 in both the index URLs and the sources below.
[[tool.uv.index]]
name = "pyg-cpu"
url = "https://data.pyg.org/whl/torch-2.8.0+cpu.html"
format = "flat"
explicit = true
[[tool.uv.index]]
name = "pyg-cu128"
url = "https://data.pyg.org/whl/torch-2.8.0+cu128.html"
format = "flat"
explicit = true

[tool.uv.sources]
torch = [
    { index = "pytorch-cpu", extra = "cpu" },
    { index = "pytorch-cu128", extra = "cu128", marker = "sys_platform == 'linux'" },
]
torch-cluster = [
    { index = "pyg-cpu", extra = "cpu" },
    { index = "pyg-cu128", extra = "cu128", marker = "sys_platform == 'linux'" },
]
torch-scatter = [
    { index = "pyg-cpu", extra = "cpu" },
    { index = "pyg-cu128", extra = "cu128", marker = "sys_platform == 'linux'" },
]
torch-sparse = [
    { index = "pyg-cpu", extra = "cpu" },
    { index = "pyg-cu128", extra = "cu128", marker = "sys_platform == 'linux'" },
]
# torch-geometric (pure-python umbrella) stays on PyPI — no index pin needed.

[tool.ruff]
line-length = 120

[tool.ruff.lint]
select = ["E", "F", "I", "UP"]
ignore = ["F722"]

[tool.ruff.lint.per-file-ignores]
"src/affex/data/types.py" = ["F821"]
"src/affex/data/transform/graph_builder.py" = ["F821"]

[tool.pytest.ini_options]
testpaths = ["tests"]

[tool.pyright]
pythonVersion = "3.12"