[PyPI] Packaging for PyPI distribution (#140)

.gitignore

@@ -7,6 +7,7 @@
 *.csv
 build/
 docs/build/
+dist/
 
 *.pkl
 *.png

MANIFEST.in

+include LICENSE
+include requirements.txt
+
+recursive-include csrc *

cacheflow/__init__.py

@@ -5,6 +5,8 @@ from cacheflow.server.arg_utils import ServerArgs
 from cacheflow.server.llm_server import LLMServer
 from cacheflow.server.ray_utils import initialize_cluster
 
+__version__ = "0.1.0"
+
 __all__ = [
     "LLM",
     "SamplingParams",

docs/source/getting_started/installation.rst

@@ -32,7 +32,7 @@ You can install CacheFlow using pip:
     $ conda activate myenv
 
     $ # Install CacheFlow.
-    $ pip install cacheflow
+    $ pip install cacheflow  # This may take 5-10 minutes.
 
 
 .. _build_from_source:
@@ -46,5 +46,4 @@ You can also build and install CacheFlow from source.
 
     $ git clone https://github.com/WoosukKwon/cacheflow.git
     $ cd cacheflow
-    $ pip install -r requirements.txt
     $ pip install -e .  # This may take 5-10 minutes.

pyproject.toml

+[build-system]
+requires = [
+    "ninja",
+    "packaging",
+    "setuptools",
+    "torch >= 2.0.0",
+    "wheel",
+]
+build-backend = "setuptools.build_meta"

setup.py

+import io
+import os
+import re
 import subprocess
 from typing import List, Set
 
 from packaging.version import parse, Version
 import setuptools
 import torch
-from torch.utils.cpp_extension import BuildExtension, CUDAExtension
-from torch.utils.cpp_extension import CUDA_HOME
+from torch.utils.cpp_extension import BuildExtension, CUDAExtension, CUDA_HOME
+
+ROOT_DIR = os.path.dirname(__file__)
 
 # Compiler flags.
-CXX_FLAGS = ["-g", "-O2"]
+CXX_FLAGS = ["-g", "-O2", "-std=c++17"]
 # TODO(woosuk): Should we use -O3?
-NVCC_FLAGS = ["-O2"]
+NVCC_FLAGS = ["-O2", "-std=c++17"]
 
+ABI = 1 if torch._C._GLIBCXX_USE_CXX11_ABI else 0
+CXX_FLAGS += [f"-D_GLIBCXX_USE_CXX11_ABI={ABI}"]
+NVCC_FLAGS += [f"-D_GLIBCXX_USE_CXX11_ABI={ABI}"]
 
 if not torch.cuda.is_available():
     raise RuntimeError(
@@ -102,15 +109,58 @@ activation_extension = CUDAExtension(
 ext_modules.append(activation_extension)
 
 
+def get_path(*filepath) -> str:
+    return os.path.join(ROOT_DIR, *filepath)
+
+
+def find_version(filepath: str):
+    """Extract version information from the given filepath.
+
+    Adapted from https://github.com/ray-project/ray/blob/0b190ee1160eeca9796bc091e07eaebf4c85b511/python/setup.py
+    """
+    with open(filepath) as fp:
+        version_match = re.search(
+            r"^__version__ = ['\"]([^'\"]*)['\"]", fp.read(), re.M)
+        if version_match:
+            return version_match.group(1)
+        raise RuntimeError("Unable to find version string.")
+
+
+def read_readme() -> str:
+    """Read the README file."""
+    return io.open(get_path("README.md"), "r", encoding="utf-8").read()
+
+
 def get_requirements() -> List[str]:
     """Get Python package dependencies from requirements.txt."""
-    with open("requirements.txt") as f:
+    with open(get_path("requirements.txt")) as f:
         requirements = f.read().strip().split("\n")
     return requirements
 
 
 setuptools.setup(
     name="cacheflow",
+    version=find_version(get_path("cacheflow", "__init__.py")),
+    author="CacheFlow Team",
+    author_email="cacheflow@gmail.com",
+    license="Apache 2.0",
+    description="CacheFlow: A high-performance LLM Serving System",
+    long_description=read_readme(),
+    long_description_content_type="text/markdown",
+    url="https://github.com/WoosukKwon/cacheflow",
+    project_urls={
+        "Homepage": "https://github.com/WoosukKwon/cacheflow",
+        "Documentation": "https://cacheflow.readthedocs.io/en/latest/",
+    },
+    classifiers=[
+        "Programming Language :: Python :: 3.8",
+        "Programming Language :: Python :: 3.9",
+        "Programming Language :: Python :: 3.10",
+        "License :: OSI Approved :: Apache Software License",
+        "Topic :: Scientific/Engineering :: Artificial Intelligence",
+    ],
+    packages=setuptools.find_packages(
+        exclude=("benchmarks", "csrc", "docs", "examples", "tests")),
     python_requires=">=3.8",
     install_requires=get_requirements(),
     ext_modules=ext_modules,