pyproject.toml 4.67 KB
Newer Older
1
[build-system]
2
# Should be mirrored in requirements/build.txt
3
requires = [
4
    "cmake>=3.26",
5
    "ninja",
6
7
    "packaging>=24.2",
    "setuptools>=77.0.3,<80.0.0",
8
    "setuptools-scm>=8.0",
Huy Do's avatar
Huy Do committed
9
    "torch == 2.7.0",
10
    "wheel",
11
    "jinja2",
12
13
]
build-backend = "setuptools.build_meta"
14

15
16
17
[project]
name = "vllm"
authors = [{name = "vLLM Team"}]
18
19
license = "Apache-2.0"
license-files = ["LICENSE"]
20
21
22
23
24
25
26
27
28
29
30
31
32
readme = "README.md"
description = "A high-throughput and memory-efficient inference and serving engine for LLMs"
classifiers = [
    "Programming Language :: Python :: 3.9",
    "Programming Language :: Python :: 3.10",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
    "Intended Audience :: Developers",
    "Intended Audience :: Information Technology",
    "Intended Audience :: Science/Research",
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
    "Topic :: Scientific/Engineering :: Information Analysis",
]
Martin Hoyer's avatar
Martin Hoyer committed
33
requires-python = ">=3.9,<3.13"
34
35
36
37
dynamic = [ "version", "dependencies", "optional-dependencies"]

[project.urls]
Homepage="https://github.com/vllm-project/vllm"
38
39
Documentation="https://docs.vllm.ai/en/latest/"
Slack="https://slack.vllm.ai/"
40
41
42
43

[project.scripts]
vllm = "vllm.entrypoints.cli.main:main"

44
45
46
[project.entry-points."vllm.general_plugins"]
lora_filesystem_resolver = "vllm.plugins.lora_resolvers.filesystem_resolver:register_filesystem_resolver"

47
[tool.setuptools_scm]
48
# no extra settings needed, presence enables setuptools-scm
49
50
51

[tool.setuptools.packages.find]
where = ["."]
52
include = ["vllm*"]
53

54
55
[tool.yapfignore]
ignore_patterns = [
56
    ".buildkite/**",
57
    "benchmarks/**",
58
59
60
    "build/**",
]

61
62
63
[tool.ruff]
# Allow lines to be as long as 80.
line-length = 80
64
65
exclude = [
    # External file, leaving license intact
66
67
    "examples/other/fp8/quantizer/quantize.py",
    "vllm/vllm_flash_attn/flash_attn_interface.pyi"
68
]
69

70
[tool.ruff.lint.per-file-ignores]
71
"vllm/third_party/**" = ["ALL"]
72
73
"vllm/version.py" = ["F401"]
"vllm/_version.py" = ["ALL"]
74
# Python 3.8 typing - skip V0 code
75
76
77
78
79
80
81
"vllm/attention/**/*.py" = ["UP006", "UP035"]
"vllm/core/**/*.py" = ["UP006", "UP035"]
"vllm/engine/**/*.py" = ["UP006", "UP035"]
"vllm/executor/**/*.py" = ["UP006", "UP035"]
"vllm/prompt_adapter/**/*.py" = ["UP006", "UP035"]
"vllm/spec_decode/**/*.py" = ["UP006", "UP035"]
"vllm/worker/**/*.py" = ["UP006", "UP035"]
82
# Python 3.8 typing - skip utils for ROCm
83
"vllm/utils.py" = ["UP006", "UP035"]
84

85
86
87
88
89
90
91
[tool.ruff.lint]
select = [
    # pycodestyle
    "E",
    # Pyflakes
    "F",
    # pyupgrade
92
    "UP",
93
94
95
96
97
98
    # flake8-bugbear
    "B",
    # flake8-simplify
    "SIM",
    # isort
    # "I",
99
    # flake8-logging-format
100
    "G",
101
102
103
104
105
106
]
ignore = [
    # star imports
    "F405", "F403",
    # lambda expression assignment
    "E731",
107
108
    # Loop control variable not used within loop body
    "B007",
109
110
    # f-string format
    "UP032",
111
112
    # Can remove once 3.10+ is the minimum Python version
    "UP007",
113
]
114
115
116

[tool.mypy]
ignore_missing_imports = true
117
check_untyped_defs = true
118
follow_imports = "silent"
119

120
# After fixing type errors resulting from follow_imports: "skip" -> "silent",
121
# move the directory here and remove it from tools/mypy.sh
122
123
124
125
files = [
    "vllm/*.py",
    "vllm/adapter_commons",
    "vllm/assets",
126
    "vllm/entrypoints",
127
    "vllm/core",
128
    "vllm/inputs",
129
    "vllm/logging_utils",
130
131
132
133
134
135
    "vllm/multimodal",
    "vllm/platforms",
    "vllm/transformers_utils",
    "vllm/triton_utils",
    "vllm/usage",
]
136
# TODO(woosuk): Include the code from Megatron and HuggingFace.
137
138
exclude = [
    "vllm/model_executor/parallel_utils/|vllm/model_executor/models/",
139
140
    # Ignore triton kernels in ops.
    'vllm/attention/ops/.*\.py$'
141
]
142
143

[tool.codespell]
144
ignore-words-list = "dout, te, indicies, subtile, ElementE"
145
skip = "tests/models/fixtures/*,tests/prompts/*,benchmarks/sonnet.txt,tests/lora/data/*,build/*,vllm/third_party/*"
146
147

[tool.isort]
148
149
150
151
skip_glob = [
    ".buildkite/*",
    "benchmarks/*",
]
152
153
use_parentheses = true
skip_gitignore = true
154
155
156
157

[tool.pytest.ini_options]
markers = [
    "skip_global_cleanup",
158
159
    "core_model: enable this model test in each PR instead of only nightly",
    "cpu_model: enable this model test in CPU tests",
160
161
    "split: run this test as part of a split",
    "distributed: run this test only in distributed GPU tests",
Joe Runde's avatar
Joe Runde committed
162
    "skip_v1: do not run this test with v1",
163
    "optional: optional tests that are automatically skipped, include --optional to run them",
164
]
165
166

[tool.pymarkdown]
167
plugins.md004.style = "sublist" # ul-style
168
169
170
171
plugins.md013.enabled = false # line-length
plugins.md041.enabled = false # first-line-h1
plugins.md033.enabled = false # inline-html
plugins.md024.allow_different_nesting = true # no-duplicate-headers
Aaron Pham's avatar
Aaron Pham committed
172
173
174
175
176
177

[tool.ty]
respect-ignore-files = true

[tool.ty.environment]
python = "./.venv"