pyproject.toml

[project]
name = "nanotron"
version = "0.4"
description = "Minimalistic Large Language Model Training and Finetuning"
authors = [
    {name = "Nouamane Tazi", email="nouamane@huggingface.co"},
    {name = "Thomas Wang", email="thomas.wang@huggingface.co"},
    {name = "Kunhao Zheng", email="kunhao@huggingface.co"},
    {name = "Thomas Wolf", email="thomas@huggingface.co"},
]
readme = "README.md"
requires-python = "~=3.10"
classifiers = [
    "Topic :: Software Development"
]
dependencies = [
    "torch>=1.13.1",
    "pyyaml",
    "numpy<2",
    "packaging",
    "safetensors",
    "dacite",
    "tqdm",
    "datasets",
    "torchtyping"
]

[tool.setuptools.packages.find]
where = ["src"]  # list of folders that contain the packages (["."] by default)

[tool.ruff]
line-length = 119
ignore = ["C901","E501"] # E501 ignores length violation handled by black
select = ["C","E","F","I","W"]
ignore-init-module-imports = true

[project.optional-dependencies]
dev = [
    "pre-commit",
    "pylint"
]

test = [
    "pytest",
    "pytest-xdist"
]

fast-modeling = [
    "flash-attn>=2.5.0,<2.7.0",
]

nanosets = [
     "transformers",
     "datatrove[io,processing]@git+https://github.com/huggingface/datatrove",
     "numba",
]

s3 = [
    "boto3",
    "s3fs",
    "s5cmd",
]

[build-system]
requires = [
    "setuptools",
]

[pytest]
norecursedirs="tests/helpers"