torch-mlir/setup.py

270 lines
9.4 KiB
Python
Raw Normal View History

# Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
# See https://llvm.org/LICENSE.txt for license information.
# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
# Also available under a BSD-style license. See LICENSE.
# Script for generating the torch-mlir wheel.
# ```
# $ python setup.py bdist_wheel
# ```
# Environment variables you are probably interested in:
#
# TORCH_MLIR_PYTHON_PACKAGE_VERSION:
# specify the version of torch-mlir, for example, this can be "20220330.357"
# for a snapshot release on 2022-03-30 with build number 357.
#
# TORCH_MLIR_ENABLE_LTC:
# enables the Lazy Tensor Core Backend
#
# LLVM_INSTALL_DIR:
# build the project *out-of-tree* using the built llvm-project
#
# CMAKE_BUILD_TYPE:
# specify the build type: DEBUG/RelWithDebInfo/Release
#
# TORCH_MLIR_CMAKE_BUILD_DIR:
# specify the cmake build directory
#
# TORCH_MLIR_CMAKE_ALREADY_BUILT:
# the `TORCH_MLIR_CMAKE_BUILD_DIR` directory has already been compiled,
# and the CMake compilation process will not be executed again.
# On CIs, it is often advantageous to re-use/control the CMake build directory.
#
# MAX_JOBS:
# maximum number of compile jobs we should use to compile your code
#
# It is recommended to build with Ninja and ccache. To do so, set environment
# variables by prefixing to above invocations:
# ```
# CMAKE_GENERATOR=Ninja CMAKE_C_COMPILER_LAUNCHER=ccache CMAKE_CXX_COMPILER_LAUNCHER=ccache
# ```
#
# Implementation notes:
# The contents of the wheel is just the contents of the `python_packages`
# directory that our CMake build produces. We go through quite a bit of effort
# on the CMake side to organize that directory already, so we avoid duplicating
# that here, and just package up its contents.
import os
import pathlib
import shutil
import subprocess
import sys
import multiprocessing
from distutils.command.build import build as _build
from setuptools import setup, Extension
from setuptools.command.build_ext import build_ext
from setuptools.command.build_py import build_py
if "develop" in sys.argv:
print("Warning: The setup.py script is only used for building the wheel package.")
print("For initializing the development environment,"
"please use the cmake commands introduced in the docs/development.md.")
sys.exit(1)
def _check_env_flag(name: str, default=None) -> bool:
return str(os.getenv(name, default)).upper() in ["ON", "1", "YES", "TRUE", "Y"]
PACKAGE_VERSION = os.getenv("TORCH_MLIR_PYTHON_PACKAGE_VERSION", "0.0.1")
# If true, enable LTC build by default
TORCH_MLIR_ENABLE_LTC = _check_env_flag("TORCH_MLIR_ENABLE_LTC", True)
TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS = _check_env_flag(
"TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS", False
)
LLVM_INSTALL_DIR = os.getenv("LLVM_INSTALL_DIR", None)
SRC_DIR = pathlib.Path(__file__).parent.absolute()
CMAKE_BUILD_TYPE = os.getenv("CMAKE_BUILD_TYPE", "Release")
TORCH_MLIR_CMAKE_ALREADY_BUILT = _check_env_flag(
"TORCH_MLIR_CMAKE_ALREADY_BUILT", False
)
TORCH_MLIR_CMAKE_BUILD_DIR = os.getenv("TORCH_MLIR_CMAKE_BUILD_DIR")
MAX_JOBS = os.getenv("MAX_JOBS", str(multiprocessing.cpu_count()))
# Build phase discovery is unreliable. Just tell it what phases to run.
class CustomBuild(_build):
def initialize_options(self):
_build.initialize_options(self)
# Make setuptools not steal the build directory name,
# because the mlir c++ developers are quite
# used to having build/ be for cmake
self.build_base = "setup_build"
def run(self):
self.run_command("build_py")
self.run_command("build_ext")
self.run_command("build_scripts")
class CMakeBuild(build_py):
def cmake_build(self, cmake_build_dir):
llvm_dir = str(SRC_DIR / "externals" / "llvm-project" / "llvm")
cmake_config_args = [
f"cmake",
f"-DCMAKE_BUILD_TYPE={CMAKE_BUILD_TYPE}",
f"-DPython3_EXECUTABLE={sys.executable}",
f"-DPython3_FIND_VIRTUALENV=ONLY",
f"-DMLIR_ENABLE_BINDINGS_PYTHON=ON",
f"-DLLVM_TARGETS_TO_BUILD=host",
f"-DLLVM_ENABLE_ZSTD=OFF",
# Optimization options for building wheels.
f"-DCMAKE_VISIBILITY_INLINES_HIDDEN=ON",
f"-DCMAKE_C_VISIBILITY_PRESET=hidden",
f"-DCMAKE_CXX_VISIBILITY_PRESET=hidden",
f"-DTORCH_MLIR_ENABLE_LTC={'ON' if TORCH_MLIR_ENABLE_LTC else 'OFF'}",
f"-DTORCH_MLIR_ENABLE_PYTORCH_EXTENSIONS={'OFF' if TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS else 'ON'}",
]
if LLVM_INSTALL_DIR:
cmake_config_args += [
f"-DMLIR_DIR='{LLVM_INSTALL_DIR}/lib/cmake/mlir/'",
f"-DLLVM_DIR='{LLVM_INSTALL_DIR}/lib/cmake/llvm/'",
f"{SRC_DIR}",
]
else:
cmake_config_args += [
f"-DLLVM_ENABLE_PROJECTS=mlir",
f"-DLLVM_EXTERNAL_PROJECTS='torch-mlir'",
f"-DLLVM_EXTERNAL_TORCH_MLIR_SOURCE_DIR={SRC_DIR}",
f"{llvm_dir}",
]
cmake_build_args = [
f"cmake",
f"--build",
f".",
f"--config",
f"{CMAKE_BUILD_TYPE}",
f"--target",
f"TorchMLIRPythonModules",
f"--",
f"-j{MAX_JOBS}",
]
try:
subprocess.check_call(cmake_config_args, cwd=cmake_build_dir)
subprocess.check_call(cmake_build_args, cwd=cmake_build_dir)
except subprocess.CalledProcessError as e:
print("cmake build failed with\n", e)
print("debug by follow cmake command:")
sys.exit(e.returncode)
finally:
print(f"cmake config: {' '.join(cmake_config_args)}")
print(f"cmake build: {' '.join(cmake_build_args)}")
print(f"cmake workspace: {cmake_build_dir}")
def run(self):
target_dir = self.build_lib
cmake_build_dir = TORCH_MLIR_CMAKE_BUILD_DIR
if not cmake_build_dir:
cmake_build_dir = os.path.abspath(
os.path.join(target_dir, "..", "cmake_build")
)
if LLVM_INSTALL_DIR:
python_package_dir = os.path.join(
cmake_build_dir, "python_packages", "torch_mlir"
)
else:
python_package_dir = os.path.join(
cmake_build_dir, "tools", "torch-mlir", "python_packages", "torch_mlir"
)
if not TORCH_MLIR_CMAKE_ALREADY_BUILT:
os.makedirs(cmake_build_dir, exist_ok=True)
cmake_cache_file = os.path.join(cmake_build_dir, "CMakeCache.txt")
if os.path.exists(cmake_cache_file):
os.remove(cmake_cache_file)
# NOTE: With repeated builds for different Python versions, the
# prior version binaries will continue to accumulate. IREE uses
# a separate install step and cleans the install directory to
# keep this from happening. That is the most robust. Here we just
# delete the directory where we build native extensions to keep
# this from happening but still take advantage of most of the
# build cache.
mlir_libs_dir = os.path.join(python_package_dir, "torch_mlir", "_mlir_libs")
if os.path.exists(mlir_libs_dir):
print(f"Removing _mlir_mlibs dir to force rebuild: {mlir_libs_dir}")
shutil.rmtree(mlir_libs_dir)
else:
print(f"Not removing _mlir_libs dir (does not exist): {mlir_libs_dir}")
self.cmake_build(cmake_build_dir)
if os.path.exists(target_dir):
shutil.rmtree(target_dir, ignore_errors=False, onerror=None)
shutil.copytree(python_package_dir, target_dir, symlinks=False)
class CMakeExtension(Extension):
def __init__(self, name, sourcedir=""):
Extension.__init__(self, name, sources=[])
self.sourcedir = os.path.abspath(sourcedir)
class NoopBuildExtension(build_ext):
def build_extension(self, ext):
pass
2022-12-07 23:01:55 +08:00
with open("README.md", "r", encoding="utf-8") as fh:
long_description = fh.read()
Upstream the ONNX importer. (#2636) This is part 1 of 2, which will also include upstreaming the FX importer. I started with ONNX because it forces some project layout updates and is more self contained/easier as a first step. Deviating somewhat from the RFCs on project layout, I made the following decisions: * Locating the `onnx_importer.py` into `torch_mlir.extras` as Maks already has opened up that namespace and it seemed to fit. Better to have fewer things at that level. * Setup the build so that the root project only contains MLIR Python and pure Python deps (like the importers), but this can be augmented with the `projects/` adding more depending on which features are enabled. * The default build continues to build everything whereas in `TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS=1` mode, it builds a `torch-mlir-core` wheel with the pure contents only. `onnx_importer.py` and `importer_smoke_test.py` are almost verbatim copies from SHARK-Turbine. I made some minor local alterations to adapt to paths and generalize the way they interact with the outer project. I expect I can copy these back to Turbine verbatim from here. I also updated the license boilerplate (they have the same license but slightly different project norms for the headers) but retained the correct copyright. Other updates: * Added the ONNX importer unit test (which also can generate test data) in lit, conditioned on the availability of the Python `onnx` package. In a followup once I know everything is stable, I'll add another env var that the CI can set to always enable this so we know conclusively if tests pass. * Moved the ONNX conversion readme to `docs/`. * Renamed CMake option `TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS` -> `TORCH_MLIR_ENABLE_PYTORCH_EXTENSIONS` and inverted the sense. Made the JitIR importer and LTC options `cmake_dependent_options` for robustness.
2023-12-13 11:02:51 +08:00
# Requires and extension modules depend on whether building PyTorch
# extensions.
INSTALL_REQUIRES = [
"numpy",
"packaging",
]
EXT_MODULES = [
CMakeExtension("torch_mlir._mlir_libs._torchMlir"),
]
NAME = "torch-mlir-core"
# If building PyTorch extensions, customize.
if not TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS:
import torch
NAME = "torch-mlir"
INSTALL_REQUIRES.extend([
f"torch=={torch.__version__}".split("+", 1)[0],
])
EXT_MODULES.extend([
CMakeExtension("torch_mlir._mlir_libs._jit_ir_importer"),
])
setup(
Upstream the ONNX importer. (#2636) This is part 1 of 2, which will also include upstreaming the FX importer. I started with ONNX because it forces some project layout updates and is more self contained/easier as a first step. Deviating somewhat from the RFCs on project layout, I made the following decisions: * Locating the `onnx_importer.py` into `torch_mlir.extras` as Maks already has opened up that namespace and it seemed to fit. Better to have fewer things at that level. * Setup the build so that the root project only contains MLIR Python and pure Python deps (like the importers), but this can be augmented with the `projects/` adding more depending on which features are enabled. * The default build continues to build everything whereas in `TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS=1` mode, it builds a `torch-mlir-core` wheel with the pure contents only. `onnx_importer.py` and `importer_smoke_test.py` are almost verbatim copies from SHARK-Turbine. I made some minor local alterations to adapt to paths and generalize the way they interact with the outer project. I expect I can copy these back to Turbine verbatim from here. I also updated the license boilerplate (they have the same license but slightly different project norms for the headers) but retained the correct copyright. Other updates: * Added the ONNX importer unit test (which also can generate test data) in lit, conditioned on the availability of the Python `onnx` package. In a followup once I know everything is stable, I'll add another env var that the CI can set to always enable this so we know conclusively if tests pass. * Moved the ONNX conversion readme to `docs/`. * Renamed CMake option `TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS` -> `TORCH_MLIR_ENABLE_PYTORCH_EXTENSIONS` and inverted the sense. Made the JitIR importer and LTC options `cmake_dependent_options` for robustness.
2023-12-13 11:02:51 +08:00
name=NAME,
version=f"{PACKAGE_VERSION}",
author="Sean Silva",
author_email="silvasean@google.com",
description="First-class interop between PyTorch and MLIR",
2022-12-07 23:01:55 +08:00
long_description=long_description,
long_description_content_type="text/markdown",
include_package_data=True,
cmdclass={
"build": CustomBuild,
"built_ext": NoopBuildExtension,
"build_py": CMakeBuild,
},
Upstream the ONNX importer. (#2636) This is part 1 of 2, which will also include upstreaming the FX importer. I started with ONNX because it forces some project layout updates and is more self contained/easier as a first step. Deviating somewhat from the RFCs on project layout, I made the following decisions: * Locating the `onnx_importer.py` into `torch_mlir.extras` as Maks already has opened up that namespace and it seemed to fit. Better to have fewer things at that level. * Setup the build so that the root project only contains MLIR Python and pure Python deps (like the importers), but this can be augmented with the `projects/` adding more depending on which features are enabled. * The default build continues to build everything whereas in `TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS=1` mode, it builds a `torch-mlir-core` wheel with the pure contents only. `onnx_importer.py` and `importer_smoke_test.py` are almost verbatim copies from SHARK-Turbine. I made some minor local alterations to adapt to paths and generalize the way they interact with the outer project. I expect I can copy these back to Turbine verbatim from here. I also updated the license boilerplate (they have the same license but slightly different project norms for the headers) but retained the correct copyright. Other updates: * Added the ONNX importer unit test (which also can generate test data) in lit, conditioned on the availability of the Python `onnx` package. In a followup once I know everything is stable, I'll add another env var that the CI can set to always enable this so we know conclusively if tests pass. * Moved the ONNX conversion readme to `docs/`. * Renamed CMake option `TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS` -> `TORCH_MLIR_ENABLE_PYTORCH_EXTENSIONS` and inverted the sense. Made the JitIR importer and LTC options `cmake_dependent_options` for robustness.
2023-12-13 11:02:51 +08:00
ext_modules=EXT_MODULES,
python_requires=">=3.8",
Upstream the ONNX importer. (#2636) This is part 1 of 2, which will also include upstreaming the FX importer. I started with ONNX because it forces some project layout updates and is more self contained/easier as a first step. Deviating somewhat from the RFCs on project layout, I made the following decisions: * Locating the `onnx_importer.py` into `torch_mlir.extras` as Maks already has opened up that namespace and it seemed to fit. Better to have fewer things at that level. * Setup the build so that the root project only contains MLIR Python and pure Python deps (like the importers), but this can be augmented with the `projects/` adding more depending on which features are enabled. * The default build continues to build everything whereas in `TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS=1` mode, it builds a `torch-mlir-core` wheel with the pure contents only. `onnx_importer.py` and `importer_smoke_test.py` are almost verbatim copies from SHARK-Turbine. I made some minor local alterations to adapt to paths and generalize the way they interact with the outer project. I expect I can copy these back to Turbine verbatim from here. I also updated the license boilerplate (they have the same license but slightly different project norms for the headers) but retained the correct copyright. Other updates: * Added the ONNX importer unit test (which also can generate test data) in lit, conditioned on the availability of the Python `onnx` package. In a followup once I know everything is stable, I'll add another env var that the CI can set to always enable this so we know conclusively if tests pass. * Moved the ONNX conversion readme to `docs/`. * Renamed CMake option `TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS` -> `TORCH_MLIR_ENABLE_PYTORCH_EXTENSIONS` and inverted the sense. Made the JitIR importer and LTC options `cmake_dependent_options` for robustness.
2023-12-13 11:02:51 +08:00
install_requires=INSTALL_REQUIRES,
extras_require={
"onnx": [
"onnx>=1.15.0",
],
},
entry_points={
"console_scripts": [
"torch-mlir-import-onnx = torch_mlir.tools.import_onnx:_cli_main",
],
},
zip_safe=False,
)