pyproject.toml

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

[project]
name = "ludwig"
dynamic = ["version"]
description = "Declarative machine learning: End-to-end machine learning pipelines using data-driven configurations."
readme = "README.md"
license = "Apache-2.0"
requires-python = ">=3.12"
authors = [
    { name = "Piero Molino", email = "piero.molino@gmail.com" },
]
keywords = [
    "ludwig",
    "deep learning",
    "machine learning",
    "natural language processing",
    "computer vision",
]
classifiers = [
    "Development Status :: 4 - Beta",
    "Intended Audience :: Developers",
    "Intended Audience :: Science/Research",
    "License :: OSI Approved :: Apache Software License",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.12",
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
]
dependencies = [
    "numpy>=1.24",
    "pandas>=2.0",
    "scipy>=1.10",
    "tabulate>=0.9",
    "scikit-learn>=1.3",
    "tqdm>=4.60",
    "torch>=2.11",
    "torchaudio>=2.11",
    "torchcodec>=0.1",
    "torchvision>=0.26",
    "transformers>=5.0",
    "sentencepiece>=0.2",
    "spacy>=2.3",
    "PyYAML>=6.0",
    "absl-py",
    "kaggle",
    "requests>=2.28",
    "py-cpuinfo",
    "fsspec[http]",
    "dataclasses-json",
    "jsonschema>=4.17",
    "tensorboard",
    "torchmetrics>=1.0",
    "torchinfo",
    "filelock",
    "psutil",
    "protobuf>=4.0",
    "gpustat",
    "rich>=12.4.4",
    "packaging",
    "retry",
    "sacremoses",
    "bitsandbytes>=0.44.0",
    "xlwt",
    "xlrd",
    "openpyxl",
    "pyarrow>=14.0",
    "lxml",
    "datasets",
    "safetensors>=0.4",
]

[project.optional-dependencies]
serve = [
    "uvicorn",
    "httpx",
    "fastapi",
    "python-multipart",
    "prometheus_client",
]
viz = [
    "matplotlib>=3.4",
    "seaborn",
    "hiplot",
    "ptitprince",
]
distributed = [
    "dask[dataframe]>=2026.1.2",
    "pyarrow>=14.0",
    "ray[default,data,serve,tune]>=2.9",
    "GPUtil",
    "tblib",
    "awscli",
]
hyperopt = [
    "ray[default,tune]>=2.9",
    "hyperopt",
    "future<1.0",
]
llm = [
    "sentence-transformers",
    "faiss-cpu",
    "accelerate",
    "loralib",
    "peft>=0.10.0",
    "torchao>=0.17.0",
]
explain = [
    "captum",
]
benchmarking = [
    "s3fs",
]
extra = [
    "modin[ray]",
    "predibase>=2023.10.2",
]
full = [
    "ludwig[serve,viz,distributed,hyperopt,llm,explain,benchmarking]",
]
test = [
    "ludwig[full]",
    "pytest",
    "pytest-timeout",
    "pytest-rerunfailures",
    "tifffile",
    "wget",
    "aim",
    "wandb",
    "comet_ml",
    "mlflow",
    "hpbandster",
    "ConfigSpace>=1.0",
    "ax-platform",
    "sqlalchemy",
    "bayesian-optimization",
    "flaml[blendsearch]",
    "nevergrad",
    "optuna",
    "scikit-optimize",
    "zoopt",
    "hyperopt",
    "s3fs>=2022.8.2",
]

[dependency-groups]
dev = ["pre-commit", "ruff>=0.8"]
docs = ["mkdocs", "mkdocs-material", "mkdocstrings[python]", "mike"]

[project.scripts]
ludwig = "ludwig.cli:main"

[project.urls]
Homepage = "https://github.com/ludwig-ai/ludwig"
Documentation = "https://ludwig.ai"
Repository = "https://github.com/ludwig-ai/ludwig"
"Bug Tracker" = "https://github.com/ludwig-ai/ludwig/issues"

[tool.hatch.version]
source = "regex"
path = "ludwig/globals.py"
pattern = 'LUDWIG_VERSION = "(?P<version>[^"]+)"'

[tool.hatch.build.targets.wheel]
packages = ["ludwig"]
artifacts = ["ludwig/py.typed"]

[tool.hatch.build.targets.wheel.shared-data]

[tool.hatch.build.targets.sdist]
include = [
    "ludwig/",
    "README.md",
    "LICENSE",
    "pyproject.toml",
]

[tool.ruff]
target-version = "py312"
line-length = 120
exclude = ["python/"]

[tool.ruff.lint]
select = [
    "E",    # pycodestyle errors
    "W",    # pycodestyle warnings
    "F",    # pyflakes
    "I",    # isort
    "UP",   # pyupgrade (includes UP007: Optional→X|None, Union→X|Y)
    "B",    # flake8-bugbear
    "C4",   # flake8-comprehensions
    "SIM",  # flake8-simplify
    "RUF",  # ruff-specific rules
]
ignore = [
    "E501",   # line length — handled by formatter
    "B006",   # mutable default args in function — handled separately
    "B008",   # function calls in default args
    "SIM108", # ternary operator — too aggressive
    "SIM102", # nested if — too aggressive
    "SIM117", # nested with — too aggressive
    "UP035",  # deprecated imports — too broad
    "RUF012", # mutable class attrs — too aggressive for schema classes
    "B028",   # no-explicit-stacklevel in warnings.warn — deferred cleanup
    "B904",   # raise-without-from-inside-except — deferred cleanup
    "B905",   # zip-without-explicit-strict — deferred cleanup
    "B027",   # empty-method-without-abstract-decorator — deferred cleanup
    "B007",   # unused-loop-control-variable — deferred cleanup
    "RUF059", # unused-unpacked-variable — deferred cleanup
    "RUF005", # collection-literal-concatenation — style preference
    "RUF001", # ambiguous-unicode-character-string — intentional unicode in some files
    "RUF002", # ambiguous-unicode-character-docstring — intentional
    "RUF003", # ambiguous-unicode-character-comment — intentional
    "SIM105", # suppressible-exception — try/except/pass has legitimate uses
    "B023",   # function-uses-loop-variable — too many false positives
    "SIM115", # open-file-with-context-handler — deferred cleanup
    "RUF015", # unnecessary-iterable-allocation — deferred cleanup
    "C414",   # unnecessary-double-cast-or-process — deferred cleanup
    "SIM201", # negate-equal-op — deferred cleanup
    "B024",   # abstract-base-class-without-abstract-method — deferred cleanup
    "E731",   # lambda-assignment — some lambdas are intentionally concise
    "RUF046", # unnecessary-cast-to-int — deferred cleanup
    "RUF022", # unsorted-dunder-all — thematic grouping with comments is preferred
    "SIM110", # reimplemented-builtin — deferred cleanup
    "B019",   # cached-instance-method — deferred cleanup (requires careful review)
    "SIM101", # duplicate-isinstance-call — deferred cleanup
    "SIM103", # needless-bool — deferred cleanup
    "SIM401", # if-else-block-instead-of-dict-get — deferred cleanup
    "C419",   # unnecessary-comprehension-in-call — deferred cleanup
    "UP046",  # non-pep695-generic-class — Python 3.12+ syntax, deferred
    "UP042",  # replace-str-enum — deferred cleanup
    "B003",   # assignment-to-os-environ — reviewed, safe in context manager
    "SIM113", # enumerate-for-loop — deferred cleanup
    "RUF017", # quadratic-list-summation — deferred cleanup
    "B020",   # loop-variable-overrides-iterator — fixed in automl.py
]

[tool.ruff.lint.per-file-ignores]
"tests/**" = [
    "S101",   # allow assert in tests
    "B011",   # allow assertFalse in tests
    "B017",   # allow pytest.raises(Exception) in tests
    "C408",   # dict()/list() literals fine in tests
    "C416",   # unnecessary list comprehension — tests are explicit
    "C417",   # map() usage fine in tests
    "RUF013", # implicit Optional fine in test signatures
    "SIM118", # key in dict.keys() fine in tests
    "SIM109", # multiple equality comparisons fine in tests
    "SIM212", # ternary ordering fine in tests
    "F821",   # undefined name (sometimes from notebook stubs)
    "F811",   # redefinition fine in tests
    "E721",   # type comparison fine in tests
    "UP031",  # %-format fine in tests
    "RUF043", # regex metacharacter warning — backticks are not special in Python regex
]
"examples/**" = [
    "E402",   # imports not at top — notebooks interleave imports with code
    "F821",   # undefined names in notebooks (seaborn loaded earlier)
    "C408",   # dict() literals fine in examples
    "C405",   # list literal for set fine in examples
    "RUF007", # itertools.pairwise — style preference, not required
    "RUF043", # regex metacharacters in match= — intentional in examples
    "B015",   # pointless comparison fine in examples
    "SIM118", # key in dict.keys() fine in examples
    "SIM212", # ternary ordering fine in examples
    "UP031",  # %-format fine in examples
]

[tool.ruff.lint.isort]
known-first-party = ["ludwig"]
force-sort-within-sections = false
order-by-type = false

[tool.ruff.format]
quote-style = "double"
indent-style = "space"