remove_grad_op_and_kernel.py

# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This script simply removes grad ops and kernels. You should use this script
when cmake ON_INFER=ON, which can greatly reduce the volume of the inference library.
"""

import argparse
import glob
import os
import re

import reduce_lib_size_util


def parse_args():
    """Parse input arguments."""
    parser = argparse.ArgumentParser(description='Remove grad op and kernels.')
    parser.add_argument('--only_kernel', action='store_true', default=False)
    parser.add_argument('--dry_run', action='store_true', default=False)

    args = parser.parse_args()
    return args


def find_type_files(cur_dir, file_type, file_list=[]):
    next_level_dirs = os.listdir(cur_dir)
    for next_level_name in next_level_dirs:
        next_level_dir = os.path.join(cur_dir, next_level_name)
        if os.path.isfile(next_level_dir):
            if os.path.splitext(next_level_dir)[1] == file_type:
                file_list.append(next_level_dir)
        elif os.path.isdir(next_level_dir):
            find_type_files(next_level_dir, file_type, file_list)
    return file_list


def remove_grad_op_and_kernel(content, pattern1, pattern2):
    res = []
    first_match = re.findall(pattern1, content, flags=re.DOTALL)
    for match in first_match:
        res.extend(re.findall(pattern2, match, flags=re.DOTALL))
    return res, len(res)


def update_operator_cmake(cmake_file):
    """Update operator cmake.
    Args:
        cmake_file (str): cmake file path.
    """
    pat1 = 'add_subdirectory(optimizers)'
    pat2 = r'register_operators\(EXCLUDES.*?py_func_op.*?\)'

    code1 = 'if(ON_INFER)\nadd_subdirectory(optimizers)\nendif()'
    code2 = 'if(ON_INFER)\nfile(GLOB LOSS_OPS RELATIVE "${CMAKE_CURRENT_SOURCE_DIR}" "*loss_op.cc")\nstring(REPLACE ".cc" "" LOSS_OPS "${LOSS_OPS}")\nendif()'

    with open(cmake_file, 'r') as f:
        content = ''.join(f.readlines())
        content = content.replace(pat1, code1)

        match = re.findall(pat2, content, flags=re.DOTALL)
        content = content.replace(
            match[0],
            code2
            + '\n'
            + match[0].replace('py_func_op', 'py_func_op ${LOSS_OPS}'),
        )

    with open(cmake_file, 'w') as f:
        f.write(content)


if __name__ == '__main__':
    args = parse_args()

    tool_dir = os.path.dirname(os.path.abspath(__file__))

    all_op = glob.glob(
        os.path.join(tool_dir, '../paddle/fluid/operators/**/*.cc'),
        recursive=True,
    )
    all_op += glob.glob(
        os.path.join(tool_dir, '../paddle/fluid/operators/**/*.cu'),
        recursive=True,
    )

    spec_ops = ['activation_op.cc']

    (
        register_op_count,
        register_op_cpu_kernel_count,
        register_op_cuda_kernel_count,
        register_op_xpu_kernel_count,
    ) = (0, 0, 0, 0)
    register_op_kernel_count, register_op_kernel_with_custom_type_count = 0, 0

    # 1. remove all grad op and kernel
    for op_file in all_op:
        # remove all grad op
        op_pattern1 = r'REGISTER_OPERATOR\(.*?\);?'
        op_pattern2 = r'REGISTER_OPERATOR\(.*?_grad,.*?\);?'
        if args.only_kernel:
            op_pattern1 = 'DISABLE_REMOVE_GRAD_OP_' + op_pattern1
            op_pattern2 = 'DISABLE_REMOVE_GRAD_OP_' + op_pattern2

        # remove all cpu grad kernel
        cpu_kernel_pattern1 = r'REGISTER_OP_CPU_KERNEL\(.*?\);?|REGISTER_OP_CPU_KERNEL_FUNCTOR\(.*?\);?'
        cpu_kernel_pattern2 = r'REGISTER_OP_CPU_KERNEL\(.*?_grad,.*?\);?|REGISTER_OP_CPU_KERNEL_FUNCTOR\(.*?_grad,.*?\);?'

        # remove all gpu grad kernel
        gpu_kernel_pattern1 = r'REGISTER_OP_CUDA_KERNEL\(.*?\);?|REGISTER_OP_CUDA_KERNEL_FUNCTOR\(.*?\);?'
        gpu_kernel_pattern2 = r'REGISTER_OP_CUDA_KERNEL\(.*?_grad,.*?\);?|REGISTER_OP_CUDA_KERNEL_FUNCTOR\(.*?_grad,.*?\);?'

        # remove all xpu grad kernel
        xpu_kernel_pattern1 = r'REGISTER_OP_XPU_KERNEL\(.*?\);?'
        xpu_kernel_pattern2 = r'REGISTER_OP_XPU_KERNEL\(.*?_grad,.*?\);?'

        # remove custom grad kernel, mkldnn or cudnn etc.
        op_kernel_pattern1 = r'REGISTER_OP_KERNEL\(.*?\);?'
        op_kernel_pattern2 = r'REGISTER_OP_KERNEL\(.*?_grad,.*?\);?'

        custom_pattern1 = r'REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE\(.*?\);?'
        custom_pattern2 = (
            r'REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE\(.*?_grad,.*?\);?'
        )

        op_name = os.path.split(op_file)[1]
        if op_name in spec_ops:
            op_pattern1 = op_pattern1[:-1]
            op_pattern2 = op_pattern2[:-1]
            cpu_kernel_pattern1 = cpu_kernel_pattern1[:-1]
            cpu_kernel_pattern2 = cpu_kernel_pattern2[:-1]
            gpu_kernel_pattern1 = gpu_kernel_pattern1[:-1]
            gpu_kernel_pattern2 = gpu_kernel_pattern2[:-1]
            xpu_kernel_pattern1 = xpu_kernel_pattern1[:-1]
            xpu_kernel_pattern2 = xpu_kernel_pattern2[:-1]
            op_kernel_pattern1 = op_kernel_pattern1[:-1]
            op_kernel_pattern2 = op_kernel_pattern2[:-1]
            custom_pattern1 = custom_pattern1[:-1]
            custom_pattern2 = custom_pattern2[:-1]

        all_matches = []
        with open(op_file, 'r', encoding='utf-8') as f:
            content = ''.join(f.readlines())

            op, op_count = remove_grad_op_and_kernel(
                content, op_pattern1, op_pattern2
            )
            cpu_kernel, cpu_kernel_count = remove_grad_op_and_kernel(
                content, cpu_kernel_pattern1, cpu_kernel_pattern2
            )
            gpu_kernel, gpu_kernel_count = remove_grad_op_and_kernel(
                content, gpu_kernel_pattern1, gpu_kernel_pattern2
            )
            xpu_kernel, xpu_kernel_count = remove_grad_op_and_kernel(
                content, xpu_kernel_pattern1, xpu_kernel_pattern2
            )
            op_kernel, op_kernel_count = remove_grad_op_and_kernel(
                content, op_kernel_pattern1, op_kernel_pattern2
            )
            custom_kernel, custom_kernel_count = remove_grad_op_and_kernel(
                content, custom_pattern1, custom_pattern2
            )

            register_op_count += op_count
            register_op_cpu_kernel_count += cpu_kernel_count
            register_op_cuda_kernel_count += gpu_kernel_count
            register_op_xpu_kernel_count += xpu_kernel_count
            register_op_kernel_count += op_kernel_count
            register_op_kernel_with_custom_type_count += custom_kernel_count

            all_matches.extend(op)
            all_matches.extend(cpu_kernel)
            all_matches.extend(gpu_kernel)
            all_matches.extend(xpu_kernel)
            all_matches.extend(op_kernel)
            all_matches.extend(custom_kernel)

        for to_remove in all_matches:
            content = content.replace(to_remove, '')
            if args.dry_run:
                print(op_file, to_remove)

        if not args.dry_run:
            with open(op_file, 'w', encoding='utf-8') as f:
                f.write(content)

    # 2. update operators/CMakeLists.txt
    cmake_file = os.path.join(
        tool_dir, '../paddle/fluid/operators/CMakeLists.txt'
    )
    update_operator_cmake(cmake_file)

    register_pd_kernel_count = reduce_lib_size_util.remove_grad_kernels(
        args.dry_run
    )

    print('We erase all grad op and kernel for Paddle-Inference lib.')
    print('%50s%10s' % ('type', 'count'))
    print('%50s%10s' % ('REGISTER_OPERATOR', register_op_count))
    print('%50s%10s' % ('REGISTER_OP_CPU_KERNEL', register_op_cpu_kernel_count))
    print(
        '%50s%10s' % ('REGISTER_OP_CUDA_KERNEL', register_op_cuda_kernel_count)
    )
    print('%50s%10s' % ('REGISTER_OP_XPU_KERNEL', register_op_xpu_kernel_count))
    print('%50s%10s' % ('REGISTER_OP_KERNEL', register_op_kernel_count))
    print(
        '%50s%10s'
        % (
            'REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE',
            register_op_kernel_with_custom_type_count,
        )
    )
    print('%50s%10s' % ('REGISTER_OP_PD_KERNEL', register_pd_kernel_count))