pyproject.toml

[build-system]
requires = ["setuptools>=61.0"]
build-backend = "setuptools.build_meta"

[project]
name = "qserve"
version = "0.1.0"
description = "Efficient and accurate LLM serving system with W4A8KV4 quantization."
readme = "README.md"
requires-python = ">=3.9"
classifiers = [
    "Programming Language :: Python :: 3",
    "License :: OSI Approved :: Apache Software License",
]
dependencies = [
    "accelerate", "sentencepiece", "tokenizers>=0.12.1",
    "torch>=2.0.0", "torchvision", 
    "transformers==4.37.2", "datasets",
    "lm_eval==0.3.0", "texttable",
    "toml", "attributedict",
    "xformers==0.0.26.post1",
    "protobuf",
    "gradio==3.35.2", "gradio_client==0.2.9",
    "fastapi", "uvicorn",
    "pydantic==1.10.14",
    "webdataset",
]

[tool.setuptools.packages.find]
exclude = ["results*", "scripts*", "examples*"]

[tool.wheel]
exclude = ["results*", "scripts*", "examples*"]